This companion piece extends the research report Crypto Exchange TradFi Expansion: A Five-Model Taxonomy (Bitbase Research, April 23, 2026; hereafter, "the prior report"). The data cutoff of this companion is the same as the prior report: April 23, 2026.

The prior report presented five TradFi-on-crypto architectures through "architectural fingerprints," each integrating multiple dimensions — settlement, margin, regulation, distribution, and asset structure. This companion piece elevates the "counterparty risk holder" dimension within those architectural fingerprints to the primary axis, offering a second systematic reading of the same set of architectures.

The two readings are co-extensive: they describe the same set of structural arrangements, not two different sample sets. Models 1 through 5 correspond exactly across both readings, because each architectural fingerprint's multi-dimensional configuration implies a specific counterparty structure, and vice versa. The methodological value of the settlement-layer view lies in translating the abstract multi-dimensional comparison into a question closer to root cause: under stress, whose balance sheet absorbs losses first, and has that balance sheet historically absorbed losses?

This companion traces each of the five architectures defined in the prior report to a documented set of historical failure modes:

Each set of failure modes corresponds to an independent watchlist of observable signals. This companion does not predict which architecture will fail; it specifies the structural indicators that each architecture warrants monitoring against.

This companion is not a revision of the prior report. The prior report is a published research product; once published, it serves as a commitment baseline. This companion is a methodological extension of the prior report, with both standing in parallel.

The prior report defined five architectures through five independent "architectural fingerprint" presentations, each integrating settlement, margin, regulation, distribution, and asset structure across multiple dimensions — but those dimensions were not explicitly ordered or rendered upstream-downstream. This presentation left methodological space for a subsequent re-reading that elevates a single dimension to the primary axis.

This companion was prompted by a reader observation in public LinkedIn discussion in late April and early May 2026. The reader's view — that the five architectures are different answers to the question of who holds counterparty risk, and that within the prior report's architectural fingerprints, the counterparty-risk dimension does the most legible structural work in distinguishing the five architectures — sharpened our reading of our own report. We adopt this view, elevating the counterparty-risk dimension within the architectural fingerprints to the primary analytical axis. We make one calibration: we read the relationship between the settlement-layer view and the regulatory dimension as co-extensive rather than upstream-downstream; the argument is developed in Chapter 3.

We chose to publish this methodological dimension as a companion rather than fold it into a future Deep Dive or merge it into the Q4 2026 Signal Tracking report. The reason is genre purity: a future Deep Dive should carry a new theme; the Q4 Signal Tracking report should strictly address the three reverse signals defined in Chapter 9 of the prior report and not introduce new analytical dimensions; a companion is the more appropriate genre in between. Bitbase Research publishes this companion as a public fulfilment of that methodological dimension.

The five architectures in the prior report can be summarized by their architectural fingerprints as follows. Model 1 — offshore or non-securities-regulated CEXs offering USDT-settled perpetual futures (stablecoin-settled perpetuals); Model 2 — CFD brokers regulated by ESMA or comparable jurisdictions; Model 3 — off-chain-custody RWA issuers regulated by securities or trust law; Model 4 — DEX perpetual protocols with no regulatory entity at the protocol layer; Model 5 — central counterparties (CCPs) within the CFTC-licensed DCM-DCO-FCM architecture.

Re-arranged by settlement layer, the same set of architectures yields a parallel set of counterparty risk holders.

Model 1 — Stablecoin issuer plus CEX proprietary book. The user's ultimate counterparty is two-layered. First is the stablecoin issuer: its reserve composition, run-vulnerability, and connectivity to the banking system determine whether USDT can hold its 1:1 peg under stress. Second is the CEX's own proprietary book: its solvency as the counterparty to stablecoin-settled perpetual futures determines whether the user can close out and withdraw under tail events. Both counterparties have documented prior defaults specifically characterized in regulatory filings.

Model 2 — CFD broker balance sheet. Under the CFD B-book commercial model, the broker is not an intermediary but the user's direct counterparty — internalising client orders against its own balance sheet and profiting from client losses. When the jurisdictional regulator (e.g. ESMA) mandates negative-balance protection, the broker's balance sheet must absorb losses exceeding client margin under tail events; when the jurisdiction does not (e.g. Mauritius FSC), losses can pass back to clients as negative balances. This jurisdictional difference, viewed through the settlement layer, is a different allocation rule for B-book risk.

Model 3 — Off-chain custodian and transfer-agent chain. Tokenized RWA and crypto interest accounts share the same settlement-layer structure: off-chain holdings plus user-agreement terms. The off-chain holder's balance sheet, corporate governance, and Terms of Use language jointly determine whether users can recover assets in default. This structure is distributed across multiple regulated entities in the institutionalized path (e.g. BlackRock BUIDL, custodied by BNY Mellon with Securitize as the SEC-registered transfer agent); it is concentrated in a single company in the crypto-interest-account path (Celsius, BlockFi), where the Terms of Use may directly stipulate ownership transfer.

Model 4 — DEX insurance fund and auto-deleveraging (ADL) mechanism. DEX perpetual protocols have no clearing-member system in the traditional CCP sense, nor any regulated multi-layered prefunded buffer. Solvency under stress is jointly determined by the insurance fund and ADL: the insurance fund is a prefunded protocol-level reserve; when it is insufficient, ADL forcibly closes a portion of the profitable counterparty's positions to restore overall solvency. Losses are ultimately borne by protocol participants according to rules, not by an entity with a balance sheet.

Model 5 — Regulated central counterparty. The CFTC-licensed DCM-DCO-FCM architecture is the most institutionalized settlement-layer arrangement: default fund, clearing-member tiered loss-mutualization obligations, multiple layers of prefunded resources, independent governance, CPMI-IOSCO Principles for Financial Market Infrastructures (PFMI) compliance. But "most institutionalized" does not mean "stress-test-proof." When extreme market conditions exhaust the default fund within a single day, CCPs still have a documented history of activating two non-standard tools: mass cancellation and negative-pricing structural rule changes.

The loss-allocation rules of the five counterparty structures under stress are structurally non-substitutable — and that is the root cause this companion sets out to recover.

This chapter is the central methodological argument of the companion.

The initial hypothesis in the LinkedIn discussion was that the regulatory dimension is the surface (downstream) and counterparty risk is the root cause (upstream); the prior report's multi-dimensional architectural-fingerprint presentation did not render the causal priority between the two explicit. We do not adopt this upstream-downstream relationship, because it does not hold up at the level of historical institutional choice.

Consider the specific differences between the ESMA jurisdiction and the Mauritius FSC jurisdiction. ESMA adopted Decision (EU) 2018/796 on May 22, 2018 (under MiFIR Article 40), implementing — from August 1, 2018 — across the EU for retail-client CFDs: tiered leverage caps, the 50% margin close-out rule, per-account negative-balance protection, prohibition on any promotional incentive, and standardized risk warnings [1][2]. Mauritius FSC does not mandate negative-balance protection, nor a client-compensation scheme analogous to FCA FSCS [3]. This jurisdictional difference is itself a structural choice: ESMA's institutional choice mandates that B-book brokers' losses exceeding client margin under tail events be absorbed by the broker's balance sheet; Mauritius FSC's institutional choice permits those losses to pass back to clients.

In other words, "regulatory-dimension difference" and "counterparty-risk-allocation difference" are not two ends of a causal chain. They are two expressions of the same institutional choice. The DCM-DCO-FCM three-tier architecture adopted by the CFTC under Dodd-Frank Title VII (7 U.S.C. § 1a et seq., enacted 2010) is itself the institutionalization of the choice that "central counterparties hold the risk," responding to the G-20's September 2009 commitment that "all standardized OTC derivatives should be traded on exchanges or electronic trading platforms and cleared through central counterparties by the end of 2012" [4]. The CFTC's adoption in 2012 of the Legal Segregation, Operational Commingling (LSOC) rule eliminated "fellow customer risk" — if one FCM customer defaults, the DCO may not reach the assets of other non-defaulting customers [5]. This rule set is a jurisdictional choice and a counterparty-structure choice; neither can be placed "upstream" of the other.

The specific meaning of co-extensive: every counterparty structure identified in this companion corresponds directly to a particular multi-dimensional configuration of an architectural fingerprint in the prior report, and vice versa. The two readings exhibit no difference in classification boundaries; the difference is purely in reading — the settlement-layer view sits closer to the root-cause question of "whose balance sheet absorbs losses first under tail events."

Model 1's counterparty risk has two layers: the stablecoin issuer and the CEX's proprietary book. Tether's history is the most thoroughly documentable case chain at the stablecoin-issuer trust-risk layer.

The 2017 Bitfinex banking-channel rupture. Wells Fargo halted outbound USD wire transfers from Bitfinex's and Tether's four Taiwanese correspondent banks (KGI Bank, First Commercial Bank, Hwatai Commercial Bank, and Taishin Bank) at the end of March 2017 [6]. Bitfinex and Tether filed Intentional Interference with Contractual Relations litigation against Wells Fargo on April 5, 2017 in the Northern District of California (N.D. Cal.), seeking injunctive relief and damages exceeding $75,000; the suit was voluntarily dismissed on April 12, 2017 [7]. The NYAG investigation later established that, beginning in mid-2017, Tether at one point had no banking channel at all and subsequently held its cash in a trust account in the name of its general counsel; NYAG filings state the account "never exceeded $61.5 million," at a time when USDT in circulation was already in the hundreds of millions [8].

The 2018–2019 Crypto Capital missing-funds episode. NYAG filings specifically record that Bitfinex transferred approximately $850 million of commingled customer and corporate funds to Crypto Capital Corp., a Panama-registered payment processor, with no written contract; the funds were frozen by various national authorities beginning in summer 2018 [8]. Bitfinex used a "$700 million transfer from Tether reserves to the Bitfinex balance sheet" to cover the gap. NYAG opened In re James v. iFinex, Inc., Index No. 450545/2019, on April 25, 2019 [8].

The February 17, 2021 NYAG settlement. The Settlement Agreement between NYAG and iFinex/Tether took effect on February 18, 2021 and was publicly announced on February 23, 2021. Terms included an $18.5 million fine, cessation of any trading activity in New York, and quarterly disclosure of Tether reserve composition over a two-year period [8][9]. Tether and Bitfinex stated in the agreement that they "neither admit nor deny" NYAG's findings of fact.

The October 15, 2021 CFTC Order. In the Matter of Tether Holdings Limited et al. (CFTC Press Release 8450-21) and In the Matter of iFinex Inc., BFXNA Inc., and BFXWW Inc. were issued on the same day [10]. The CFTC found that, during a relevant period of at least June 1, 2016 through February 25, 2019, Tether's representations did not match its actual reserves; in a 26-month sample period, Tether's actual U.S.-dollar reserves covered USDT in circulation on only 27.6% of days; Tether's reserves contained unsecured receivables and non-fiat assets [10]. The CFTC fined Tether $41 million and Bitfinex $1.5 million [10].

Reserve-structure evolution and commercial-paper exposure. Tether's Q1 2022 attestation disclosed the composition of its $82.1 billion reserves [11]. Federal Reserve IFDP Working Paper No. 1334, Stablecoins: Growth Potential and Impact on Banking, discusses among its potential stress scenarios the transmission of stablecoin runs and the resulting forced sales of short-term credit instruments to short-term funding markets [12].

Settlement-layer conclusion. Under Model 1, the user has two ultimate counterparties: the run-vulnerability of the stablecoin issuer's reserves and the solvency of the CEX's proprietary book in perpetual futures. Both have documented prior defaults specifically characterized in regulatory filings.

Model 2's counterparty is the CFD broker itself. When a tail event exceeds client margin, the solvency of the broker's balance sheet is the root of user risk.

The January 15, 2015 Swiss-franc black-swan event. The Swiss National Bank (SNB) unilaterally announced on January 15, 2015 at 09:30 GMT the removal of the EUR/CHF floor of 1.20, in place since September 6, 2011. EUR/CHF dropped to a low of 0.7710 within minutes and closed at 1.0472 on the day, a one-day decline of 12.74%; USD/CHF dropped to 0.7462 [13].

Alpari (UK) Ltd entered Special Administration. Alpari (UK) Ltd announced insolvency on the second day after the SNB decision (January 16, 2015), and on January 19, 2015 the English High Court ruled it into the Special Administration Regime, with Richard Heis, Samantha Bewick, and Mark Firmin of KPMG LLP appointed as Joint Special Administrators [14]. KPMG reports indicated that approximately 99.8% of clients (by value) ultimately recovered funds through the Claims Portal; final distributions extended through June 2017. LeapRate reported a final client-funds shortfall of approximately $17.3 million during the final-distribution stage [15].

FXCM and the February 6, 2017 CFTC Order. FXCM incurred approximately $225 million in client negative balances on January 15, 2015 due to the franc gap, and the next day accepted a $300 million emergency loan from Leucadia [16]. Two years later, the CFTC and NFA issued settlement filings on the same day: In the Matter of Forex Capital Markets, LLC, FXCM Holdings, LLC, Dror Niv, William Ahdout (CFTC Press Release pr7528-17, February 6, 2017) [17]. The CFTC found that, between September 4, 2009 and 2014, FXCM had falsely represented a "No Dealing Desk" agency model while actually routing order flow to Effex Capital, a market maker with which FXCM had undisclosed affiliations and profit-sharing arrangements; NFA subsequently disclosed that FXCM received approximately $77 million in "order flow rebates" from Effex between 2010 and 2014 [17]. Penalties: a CFTC fine of $7 million; FXCM, FXCM Holdings, Niv, and Ahdout permanently barred from CFTC registration or any association with registered entities [17]. FXCM's U.S. customer accounts were sold to Gain Capital; the parent FXCM Inc. renamed itself Global Brokerage, Inc., and shareholders subsequently filed Shipco Transport Inc. v. Global Brokerage, Inc., Niv, Ahdout (S.D.N.Y.), ultimately settling for $6.5 million in 2023 [18].

ESMA Decision (EU) 2018/796 and B-book institutional constraint. ESMA adopted Decision (EU) 2018/796 on May 22, 2018 [1]. The accompanying ESMA35-43-1000 cited NCA survey data showing that the proportion of retail-investor CFD accounts in losing positions across major EU jurisdictions was approximately 74% to 89%; specific page numbers and table references are subject to the ESMA public document [1]. ESMA's public statement of February 2026 further explicitly brought "perpetual futures/contracts" within the scope of the existing CFD product-intervention measures [2].

Mauritius FSC as the offshore route. Mauritius FSC issues licenses for full-service investment dealers under the Securities Act 2005, Financial Services Act 2007, and AML/CFT Act 2009, permitting retail leverage typically reaching 1:500 to 1:2000; negative-balance protection is not an FSC mandatory obligation [3].

Settlement-layer conclusion. Under Model 2, the user's counterparty is the CFD broker itself. The negative-balance protection in the ESMA jurisdiction mandates that losses exceeding client margin be absorbed by the broker's balance sheet; the Mauritius FSC does not mandate this rule and losses can pass back to clients. This jurisdictional difference, viewed through the settlement layer, is a different allocation rule for B-book risk.

Model 3's counterparty risk sits off-chain. The off-chain holder's balance sheet, corporate governance, and Terms of Use language jointly determine whether users can recover assets in default.

A note: the Celsius and BlockFi cases discussed in this chapter are of the "crypto interest account" form, which differs significantly in institutional development from tokenized RWA (e.g. BlackRock BUIDL). This companion places them under Model 3 because the two share the same settlement-layer structure: off-chain holdings plus a transfer-agent chain.

Celsius Network (In re Celsius Network LLC, Bankr. S.D.N.Y. Case No. 22-10964 (MG)). Celsius Network LLC and its affiliates filed for Chapter 11 in the U.S. Bankruptcy Court for the Southern District of New York on July 13, 2022. CoinDesk reported the next day that Celsius's balance sheet had a hole of approximately $1.2 billion [19]; subsequent disclosures indicated client liabilities of approximately $4.7 billion [20]. The Earn Account was the largest product, accounting for approximately 77% of platform assets and a market value of approximately $4.2 billion.

The court appointed Shoba Pillay (a partner at Jenner & Block) as Examiner on September 29, 2022; Pillay submitted two core reports: the Interim Report (Doc 1411, November 19, 2022) and the Final Report (Doc 1956, January 31, 2023, 476 pages) [21]. The Final Report's central conclusions were: Celsius's actual operations did not match the business it had marketed and sold to customers; Celsius cumulatively spent approximately $558 million purchasing CEL in the CEL token market, and was at virtually any point the only material buyer of CEL; founder Mashinsky personally realized approximately $68.7 million through CEL sales [21].

The most decisive settlement-layer ruling came from Chief Judge Martin Glenn's Memorandum Opinion and Order Regarding Ownership of Earn Account Assets (ECF Doc. No. 1822) of January 4, 2023: based on the language in successive versions of Celsius's Terms of Use granting "Celsius … all right and title to such Digital Assets, including ownership rights," crypto assets in Earn Accounts were transferred to Celsius the moment they were deposited; from the Petition Date (July 13, 2022) onward, they became property of the bankruptcy estate [22]. Earn customers therefore became unsecured creditors. The ruling left approximately 600,000 Earn account holders without ownership of $4.2 billion in assets, with claims to be repaid only according to the bankruptcy hierarchy [22].

BlockFi (In re BlockFi Inc. et al., Bankr. D.N.J. Case No. 22-19361 (MBK)). BlockFi filed for Chapter 11 on November 28, 2022 (with its Bermuda subsidiary BlockFi International Ltd. simultaneously entering liquidation). At filing: more than 100,000 creditors; the largest unsecured creditor was Ankura Trust Company (representing BlockFi Interest Account customers) at approximately $729 million; the second largest was FTX at approximately $275 million (corresponding to part of the $400 million revolving credit facility FTX provided BlockFi in July 2022) [23]. BlockFi's bankruptcy was the first wave of FTX contagion; the legal status of customer assets was structurally identical to Celsius.

Prior to the bankruptcy, on February 14, 2022, BlockFi had reached a historic first settlement with the SEC (In the Matter of BlockFi Lending LLC, SEC Admin. Proc. File No. 3-20700, Securities Act Release No. 33-11029): BlockFi agreed to pay $100 million ($50 million to the SEC, $50 million to 32 state regulators coordinated by NASAA), and ceased selling BIA to new U.S. customers [24].

Institutionalized RWA tokenization: risk redistribution rather than elimination. Tokenized RWA distributes Model 3's settlement-layer risk across multiple parties. Take BlackRock BUIDL (launched on Ethereum in March 2024, later expanded to Polygon, Arbitrum, Optimism, Avalanche, Aptos): BlackRock Financial Management as fund manager; Securitize, LLC (SEC-registered transfer agent) handling tokenization, whitelisting, transfer agency, and issuance; The Bank of New York Mellon as custodian for cash and securities; PricewaterhouseCoopers providing independent audit [25]. BIS Bulletin No. 115 (released in 2025) noted that BUIDL and tokenized money-market funds (TMMFs) such as WisdomTree's WTGXX exhibit highly concentrated ownership — approximately 90% of BUIDL is held in four wallet addresses [26].

This structure does not eliminate Model 3's counterparty risk through the settlement-layer view; it distributes it across BNY Mellon (asset-custody failure risk), Securitize (transfer-agent record-keeping failure risk), and SPV legal wrapping (bankruptcy-isolation risk under BVI legal environments).

Settlement-layer conclusion. Under Model 3, counterparty risk sits off-chain, and the off-chain holder's balance sheet, corporate governance, Terms of Use language, and transfer-agent records are the actual determinants of whether users can recover assets. The Celsius and BlockFi cases demonstrate the bankruptcy-law outcome when "crypto interest accounts" explicitly transfer ownership at the user-agreement layer; the BUIDL-class institutionalized path demonstrates an alternative risk allocation under multi-party regulated coordination, applied to the same off-chain holding structure.

A point of explicit demarcation: crypto interest accounts and tokenized RWA share the same settlement-layer structure (off-chain holdings plus a transfer-agent chain), but they differ markedly in institutional development across governance, independent custody, transfer-agent registration, independent audit, and SPV legal wrapping. The failure mode listed in this chapter applies to the structural constraint of "explicit ownership transfer at the user-agreement layer to a single company"; the specific failure mode of the BUIDL-class institutionalized path has no documented historical precedent at present and therefore cannot be extrapolated from the Celsius/BlockFi rulings. The specific failure modes of the BUIDL-class path remain open observation items, recorded under "BVI, Cayman, Liechtenstein and other SPV legal-environment tokenized-RWA bankruptcy precedents" in the Model 3 watchlist of Chapter 9.

Under stress, Model 4's counterparty-risk-absorption capacity is determined jointly by insurance fund size, ADL trigger rules, and oracle design. The two historical events at GMX and dYdX show that an insurance fund can not only be exhausted by a smart-contract vulnerability but can also be deliberately "drained" by the protocol's own design.

Structural differences between DEX insurance funds and CCP default funds. The CPMI-IOSCO Principles for Financial Market Infrastructures (PFMI, released April 2012, with 24 principles and 5 responsibilities) require CCPs to maintain prefunded resources (the default fund) sufficient to "cover extreme but plausible market scenarios," a clear default waterfall, and recovery and orderly wind-down plans [27]. DEX-perpetual insurance funds bear a structural similarity but lack the multi-layer prefunded buffer required by PFMI, the tiered loss-mutualization obligations of clearing members, and independent governance. When the insurance fund is insufficient, DEX-perpetual protocols typically use ADL (auto-deleveraging) to forcibly close part of the profitable counterparty's positions to restore overall solvency. Tarun Chitra's arXiv submission Autodeleveraging: Impossibilities and Optimization (arXiv:2512.01112, v1 submitted November 30, 2025; v3 revised February 16, 2026) presents a trilemma theorem for ADL: no ADL strategy can simultaneously satisfy exchange solvency, revenue, and fairness to traders [28].

The September 18, 2022 GMX V1 AVAX/USD price-manipulation incident. GMX is a DEX-perpetual protocol deployed on Arbitrum and Avalanche, with a core design where LPs become market-makers by providing GLP (a basket of BTC/ETH/AVAX/stablecoin), and traders execute at Chainlink-quoted prices (zero slippage, zero price impact). Beginning at 01:15:31 UTC on September 18, 2022, a trader opened large positions across five cycles in the GMX AVAX/USD market — each at approximately $4 to 5 million — while concurrently manipulating AVAX spot prices on a centralized exchange, then closing for profit. On-chain analysis by Joshua Lim (then head of derivatives at Genesis Trading) showed the first cycle yielded approximately $158,000, with cumulative profit of approximately $565,000 (borne by GLP holders) [29]. GMX that day capped AVAX/USD long open-positions at $2 million and shorts at $1 million [29]. The event was not a smart-contract vulnerability but rather an arbitragable property of GMX's design: "zero price impact + single-oracle pricing" under large directional positions.

The November 17–18, 2023 dYdX v3 YFI insurance-fund extraction event. Beginning on November 1, 2023, YFI exhibited highly concentrated long-side accumulation — public data from dYdX founder Antonio Juliano showed that YFI open interest on dYdX v3 grew from $0.8 million to $67 million within days [30]. Between November 17 and 18, 2023, YFI's price dropped approximately 40% to 43% within hours over a single day. Attackers attempted to close positions but, due to liquidity exhaustion, mostly failed; positions ultimately auto-liquidated into negative equity, with the insurance fund absorbing the loss: dYdX v3's insurance fund paid out approximately $9 million, accounting for approximately 40% of the v3 insurance fund, leaving $13.5 million [31]. dYdX's official Post Mortem on SUSHI and YFI Incident noted that, under oracle-manipulation risk, dYdX v3's design left the insurance fund as the ultimate loss bearer; this is fundamentally different from the multi-layer waterfall of clearing members, default fund, and CCP own capital that absorbs losses in a traditional CCP default [32].

Settlement-layer conclusion. Under Model 4, the user does not face a counterparty with a balance sheet; the user faces a protocol-level loss-allocation mechanism composed of insurance-fund and ADL rules. The robustness of this mechanism depends on oracle design, liquidity parameters, and trigger rules — not on the solvency of any external entity.

Model 5 is the most institutionalized settlement-layer arrangement, but history also shows that CCPs are not "untouchable" under stress. This chapter presents two documented historical precedents: mass cancellation and negative-pricing structural rule changes.

Institutional origin: Dodd-Frank Title VII. The Dodd-Frank Wall Street Reform and Consumer Protection Act, Title VII (7 U.S.C. § 1a et seq.), was enacted in 2010 in response to the G-20's September 2009 commitment on central clearing of OTC derivatives. Its core structural product is the three-tier architecture: DCM (Designated Contract Market) as the execution venue, DCO (Derivatives Clearing Organization) as the central counterparty, and FCM (Futures Commission Merchant) as the intermediary and clearing member between the customer and the DCO [4]. The CFTC's adoption in 2012 of the LSOC rule eliminated "fellow customer risk" [5].

The March 2022 LME nickel suspension and trade-cancellation episode. The LME three-month nickel price closed at $50,300/tonne on March 4, 2022; $48,078/tonne on March 7; and surged to $101,365/tonne within hours of opening on March 8. The LME suspended trading at 08:15 GMT on March 8 and at 12:05 GMT decided to cancel all nickel trades after 00:00 GMT that day, with a total cancellation size of approximately $12 billion [33]. The market remained suspended until reopening on March 16.

On the regulatory-response side: the FCA, PRA, and Bank of England issued a joint statement initiating review on April 4, 2022 [34]; the LME commissioned an independent review by Oliver Wyman, with the Final Report released on January 10, 2023 [35]. On March 3, 2023, the Bank of England publicly announced its supervisory action on LME Clear, citing LME Clear's deficiencies in governance, management, and risk management across multiple layers, and appointing a skilled person under Section 166 of the Financial Services and Markets Act 2000 to provide long-term oversight of LME Clear's remediation [36]. The FCA on the same day announced an enforcement investigation into the LME — the first time a UK regulator publicly initiated enforcement against an exchange [37].

Settlement-layer data (based on Office of Financial Research Working Paper No. 24-09) shows: in Q1 2022, LME Clear's cumulative margin-breach amount was $23.3 billion, two orders of magnitude higher than several preceding quarters; the largest single-account breach was $2.0 billion, by itself exceeding LME Clear's $1.1 billion default fund; variation-margin calls on March 3, March 4, and March 7 contributed approximately 65% of the excess breach amount [38]. These data show that even with an institutionalized default waterfall, the largest single-day breach in the nickel event already exceeded the default fund's coverage. The LME therefore activated Trading Rule 22 to cancel executed trades — a non-standard CCP tool when the default fund is insufficient.

The judicial challenge: Elliott Associates and Jane Street Global Trading filed Judicial Review in 2022, asserting that the LME's decision violated public-law obligations under Recognised Investment Exchanges and breached the "peaceful enjoyment of property" right under Human Rights Act 1998; combined claims totaled approximately $472 million [39]. The Divisional Court dismissed all claims on November 29, 2023; the Court of Appeal upheld the ruling on October 7, 2024; the UK Supreme Court refused permission to appeal on January 29, 2025 [39][40]. The court's core reasoning: the LME enjoys expert discretion in judging "disorderly market," and the cancellation power conferred by Trading Rule 22 is a lawful market-maintenance tool compatible with MiFID II [40].

The April 20, 2020 CME WTI May futures negative-price settlement. The CME NYMEX WTI May contract (CL, May 2020) opened at $17.73/barrel on April 20, 2020, settled at the New York 2:30 p.m. price of −$37.63/barrel, and intraday touched a low of −$40.32/barrel [41]. That day was the day before contract expiry. E-mini WTI (QM) and the cash-settled WTI listed on ICE Europe also settled at −$37.63 by reference to the CL settlement price [41].

On the prior rule change: the CME issued Testing Opportunities in CME's "New Release" Environment for Negative Prices and Strikes for Certain NYMEX Energy Contracts (Clearing Advisory Chadv20-160) on April 8, 2020, having previously notified CFTC staff on April 1, 2020, with subsequent public notices on April 3, April 8, April 13, and April 15 [42]. The CFTC released Interim Staff Report on NYMEX WTI Crude Contract Trading on and around April 20, 2020 (CFTC Press Release 8315-20) in November 2020 [43].

Broker loss case: Interactive Brokers covered margin for clients holding May QM/WTI longs into negative settlement, with maximum estimated losses of approximately $109.3 million [44]. The CFTC fined Interactive Brokers $1.75 million in 2021 (CFTC Press Release 8432-21) for finding that the firm's electronic trading systems prior to April 20 were not configured to recognize negative prices and failed to enforce internal minimum margin requirements pre-trade, resulting in approximately $82.57 million in initial customer losses [45].

Settlement-layer conclusion. Model 5 is the most institutionalized settlement-layer arrangement, but history also shows that CCPs are not "untouchable" under stress: mass cancellation (LME) and negative-pricing structural rule change (CME) are two non-standard CCP tools used when the default fund is insufficient or when the contract faces physical-delivery breakdown. The settlement-layer question is not "will Model 5 fail," but "are the tools Model 5 invokes under stress anticipated and priced by market participants?"

This chapter lists the structural indicators each of the five counterparty risk holders warrants monitoring against. This is not a prediction of which architecture will fail; it specifies the observable signals corresponding to each architecture's historical failure modes. The Q4 2026 Signal Tracking report should continue to strictly address the three reverse signals defined in Chapter 9 of the prior report; the five watchlists in this chapter are new observation dimensions, which can be incorporated into ongoing tracking as supplementary signals in a future Market Insights issue after the Q4 report.

Model 1 — Stablecoin issuers. Watch: the share of commercial paper, corporate bonds, and secured loans in major stablecoin issuers' quarterly attestation reports; rolling changes in short-term Treasury composition; issuer banking-channel (USD on/off-ramp) stability announcements; primary-market redemption spreads and frequency between major stablecoins and USD. Two historical baselines for primary-market redemption pressure: USDT briefly traded at approximately $0.95 during the LUNA/UST collapse in May–June 2022 [46]; USDC briefly de-pegged to approximately $0.88 during the SVB event in March 2023, recovering within days [47].

Model 2 — CFD broker balance sheets. Watch: emergency-funding announcements by major B-book brokers following tail events; the enforcement status of negative-balance protection by jurisdictional regulators; the evolution of client-fund segregation rules in offshore jurisdictions such as Mauritius FSC and Seychelles FSA; consolidated audit disclosures of major brokers' client assets versus corporate own funds. The historical baseline is the regulatory restructuring of multiple brokers within 6 to 24 months following the 2015 SNB event.

Model 3 — Off-chain custodians and transfer-agent chains. Watch: the specific language regarding asset ownership in crypto-interest-account providers' Terms of Use; the legal status of bankruptcy-isolation declarations by off-chain custodians (BNY Mellon, State Street, etc.); the operational continuity of SEC-registered transfer agents; tokenized-RWA bankruptcy precedents under BVI, Cayman, Liechtenstein, and other SPV legal environments. The Earn-asset ruling by Judge Glenn on January 4, 2023 in the Celsius case is the historical anchor for this watchlist [22].

Model 4 — DEX insurance funds and ADL. Watch: rolling changes in insurance-fund balances of major DEX-perpetual protocols; ADL trigger frequency and post-trigger deleveraging ratios; the multi-source diversification of oracle pricing and per-oracle position limits; the ratio of long-tail asset (YFI, SUSHI-class) open interest on DEX perpetuals to the asset's spot-market float. Both the GMX V1 and dYdX v3 events occurred on long-tail assets [29][32].

Model 5 — Regulated CCPs. Watch: quarterly default-fund size and margin-breach disclosures by major CCPs (CME, ICE, LME Clear, LCH, Eurex Clearing); CCPs' contract-rule pre-disclosure of negative-price, mass-cancellation, and forced-liquidation tools; supervisory-action announcements by CCP regulators (CFTC, Bank of England PRA, ECB) on CCP governance and risk management. The Bank of England's appointment of a skilled person to oversee LME Clear's remediation on March 3, 2023 is the historical anchor for this watchlist [36].

The five watchlists do not constitute investment advice and do not constitute specific probability estimates of future events. They are structural indicators used to classify new stress events back into existing failure modes when they occur, rather than treating each new event as without precedent.

Research scope and limitations. This companion extends the five TradFi-on-crypto architectures defined in the prior report by elevating the "counterparty risk holder" dimension within the architectural fingerprints to the primary analytical axis. This report does not construct a new architectural taxonomy, does not predict which architecture will fail first or fail ultimately, does not assess specific solvency probabilities for any individual issuer or exchange under stress, and does not develop unit-economic comparison of the five architectures at the exchange level or financial modelling of any specific issuer. Each architecture's failure mode occurred under different market structures, macroeconomic environments, and regulatory maturities, and does not constitute an extrapolatable ranking of failure probabilities. This companion does not treat the FTX collapse (November 11, 2022) as a single-type representative case, because FTX's architectural hybridity (exchange + Alameda proprietary market-maker + customer-fund misappropriation) gave it partial features across Models 1, 3, and 4; FTX is mentioned in this report only in the context of the BlockFi exposure.

Data currency. All historical-case data, regulatory documents, bankruptcy court filings, official post-event reports, and official academic working papers cited in this report are based on publicly available information as of April 23, 2026. The data cutoff of this companion is the same as the prior report, in order to maintain comparability with the prior report. New events between the prior report's release and this companion's writing are not included in the analysis, in order to avoid functional conflict with the Q4 2026 Signal Tracking report. Readers should treat this report as an analysis at a specific time slice.

Research independence. This report is independently written by Bitbase Research, and its analytical conclusions are based on publicly available primary sources and the research team's independent judgment. The "five counterparty risk holders" analytical axis adopted in this report is a research construct and does not represent the official classification of any regulator or standards body. Specific institution names referenced in this report (including Tether, Bitfinex, Wells Fargo, Crypto Capital Corp., FXCM, Alpari UK, KPMG, Leucadia, Effex Capital, Gain Capital, Celsius Network, BlockFi, Ankura Trust Company, Jenner & Block, BlackRock, Securitize, BNY Mellon, PricewaterhouseCoopers, WisdomTree, Hashnote, Ondo, GMX, dYdX, SUSHI, YFI, Genesis Trading, Chainlink, LME, LME Clear, CME, CME Group, ICE, ICE Europe, LCH, Eurex Clearing, Oliver Wyman, Elliott Associates, Jane Street Global Trading, Interactive Brokers) serve only as objective references for describing the industry landscape; inclusion is not an endorsement, and absence is not a negative signal.

Conflict-of-interest disclosure. Bitbase operates a centralized exchange and may offer products within Model 1 (CEX-native stablecoin-settled perpetual futures) as analyzed in this report. Readers should take this into account when interpreting the report's analysis of Model 1 and its comparison to the other four models. The analytical framework in this report was developed independently of any specific product plan, and the report makes no statement about any specific Bitbase product, current or upcoming. The report's arguments and watchlists apply symmetrically to all five models, including the one Bitbase may occupy.

Tools and generation assistance. This report used large language models as research-assistance tools in material gathering, cross-source fact verification, structured argument, and draft writing. All primary data, regulatory documents, bankruptcy court filings, attestation reports, and market indicators were human-verified against their original sources. Specific numerical figures, direct quoted language from regulators and judges, case numbers, ECF document numbers, and Press Release numbers were manually traced to primary sources. Core claims and judgments were made independently by Bitbase Research team members; reverse self-examination paragraphs were authored by humans. We acknowledge the inherent error risk of AI-assisted research in long-tail data handling and have reduced it through multi-round fact-checking, but cannot fully eliminate it. The methodological dimension of this companion was prompted by an observation from a reader in public LinkedIn discussion in late April and early May 2026; we do not name the reader but have recorded the source and first-appearance date in our internal Editorial Note.

Non-investment-advice statement. This report does not constitute investment advice, a recommendation to buy/sell/hold any financial instrument, or a solicitation for any financial product or service. The five counterparty-risk architectures and five watchlists identified in this report are research frameworks and not portfolio-construction methodologies; they make no statement about the expected return, risk, or suitability of any product within any architecture. Readers should consult independent, licensed financial, legal, and tax advisors before acting on any information in this report. Crypto-asset trading carries substantial risk, including but not limited to market volatility, liquidity risk, technical risk, regulatory risk, and the possibility of total loss of principal.

Forward-looking-statement risk. Statements in this report concerning evolution of stablecoin issuer reserves, CFD broker jurisdictional regulatory paths, off-chain custody and transfer-agent regimes, DEX insurance fund and ADL design, and the activation of tools by regulated CCPs under stress are forward-looking and carry uncertainty. Regulatory outcomes depend on rulemaking processes, enforcement decisions, and political developments that are not under any single actor's control. Readers should treat forward-looking statements as conditional on publicly available information as of April 23, 2026 and subject to revision.

Signal tracking. The five watchlists proposed in this companion do not replace the three reverse signals defined in Chapter 9 of the prior report. Bitbase Research commits to publishing a Signal Tracking follow-up in Q4 2026, revisiting the three reverse signals defined in Chapter 9 of the prior report against then-current data. The five watchlists in this companion serve as new observation dimensions, which can be incorporated into ongoing tracking as supplementary signals in a future Market Insights issue after the Q4 report.

[1] European Securities and Markets Authority, "ESMA agrees on product intervention measures in relation to CFDs and binary options offered to retail investors," March 27, 2018 (Decision (EU) 2018/796 subsequently adopted on May 22, 2018). https://www.esma.europa.eu/press-news/esma-news/esma-agrees-prohibit-binary-options-and-restrict-cfds-protect-retail-investors

[2] European Securities and Markets Authority, Public Statement on Derivatives in Scope of the CFD Product Intervention Measures, ESMA35-243228190-8024, February 2026.

[3] Mauritius Financial Services Commission, Investment Dealer (Full Service Dealer Including Underwriting) license framework under Securities Act 2005, Financial Services Act 2007, and AML/CFT Act 2009 (public licensing framework).

[4] Cornell Law School Legal Information Institute, "Dodd-Frank: Title VII - Wall Street Transparency and Accountability." https://www.law.cornell.edu/wex/dodd-frank\_title\_vii\_-\_wall\_street\_transparency\_and\_accountability

[5] Davis Polk & Wardwell LLP, CFTC Adopts Final Rule on Protection of Cleared Swap Customer Collateral, January 2012. https://www.davispolk.com/insights/client-update/cftc-adopts-final-rule-protection-cleared-swap-customer-collateral

[6] CoinDesk, "Bitfinex Withdraws Lawsuit Against Wells Fargo," April 12, 2017. https://www.coindesk.com/markets/2017/04/12/bitfinex-withdraws-lawsuit-against-wells-fargo

[7] Bitfinex, BFXNA Inc., BFXWW Inc., Tether Holdings Limited et al. v. Wells Fargo & Company, Wells Fargo Bank N.A., Case No. 3:17-cv-01882, N.D. Cal., filed April 5, 2017; voluntarily dismissed April 12, 2017.

[8] New York State Office of the Attorney General, Settlement Agreement: In re iFinex Inc., BFXNA Inc., BFXWW Inc., Tether Holdings Limited, Tether Limited, Tether Operations Limited, Tether International Limited, Index No. 450545/2019, executed February 17, 2021. https://ag.ny.gov/sites/default/files/2021.02.17\_-*settlement\_agreement*-\_execution\_version.b-t\_signed-c2\_oag\_signed.pdf

[9] New York State Office of the Attorney General, "Attorney General James Ends Virtual Currency Trading Platform Bitfinex's Illegal Activities in New York," February 23, 2021. https://ag.ny.gov/press-release/2021/attorney-general-james-ends-virtual-currency-trading-platform-bitfinexs-illegal

[10] Commodity Futures Trading Commission, "CFTC Orders Tether and Bitfinex to Pay Fines Totaling $42.5 Million," Press Release No. 8450-21, October 15, 2021. https://www.cftc.gov/PressRoom/PressReleases/8450-21

[11] Tether Holdings Limited, Independent Auditors' Report on the Consolidated Reserves Report, MHA Cayman, March 31, 2022.

[12] Federal Reserve Board, Stablecoins: Growth Potential and Impact on Banking, International Finance Discussion Paper No. 1334, January 2022. https://www.federalreserve.gov/econres/ifdp/files/ifdp1334.pdf

[13] Nasdaq / Reuters, "Forex - Swiss franc remains sharply higher after SNB scraps rate cap," January 15, 2015. https://www.nasdaq.com/articles/forex-swiss-franc-remains-sharply-higher-after-snb-scraps-rate-cap-2015-01-15

[14] BestBrokerDeals, "Alpari Insolvency Announcement" (Joint Special Administrators appointment notice). https://bestbrokerdeals.com/forex-knowledge-base/forex-broker-regulation/alpari-insolvency-announcement/

[15] LeapRate, "Alpari UK final bankruptcy distribution results in $17.3 million client funds shortfall" (KPMG Joint Special Administrators final client distribution report). https://www.leaprate.com/forex/brokers/alpari-uk-final-bankruptcy-distribution-results-15-7-million-client-funds-shortfall/

[16] Wikipedia, "FXCM," summary entry, citing the Leucadia $300M emergency loan and $225M client negative balance loss. https://en.wikipedia.org/wiki/FXCM

[17] Commodity Futures Trading Commission, In the Matter of Forex Capital Markets, LLC, FXCM Holdings, LLC, Dror Niv, William Ahdout, CFTC Press Release pr7528-17, February 6, 2017; Finance Magnates same-day reporting https://www.financemagnates.com/forex/brokers/breaking-cftc-levies-7m-fine-fxcm-forced-quit-us-market/

[18] FX News Group, "NY Court approves settlement in securities class action brought by FXCM Inc stockholders," Shipco Transport Inc. v. Global Brokerage, Inc., Niv, Ahdout, S.D.N.Y., $6.5M class-action settlement. https://fxnewsgroup.com/forex-news/retail-forex/ny-court-approves-settlement-in-securities-class-action-brought-by-fxcm-inc-stockholders/

[19] CoinDesk, "Celsius Acknowledges $1.2B Hole in Balance Sheet," July 14, 2022. https://www.coindesk.com/business/2022/07/14/celsius-acknowledges-12b-hole-in-balance-sheet

[20] Fortune, "Celsius Network owes its customers $4.7B after declaring bankruptcy," August 19, 2022. https://fortune.com/2022/08/19/celsius-network-bankruptcy-customers-savings-court/

[21] Shoba Pillay (Examiner), Final Report of Shoba Pillay, Examiner, In re Celsius Network LLC et al., Case No. 22-10964 (MG), Bankr. S.D.N.Y., ECF Doc. No. 1956, January 31, 2023 (476 pages).

[22] Chief Judge Martin Glenn, Memorandum Opinion and Order Regarding Ownership of Earn Account Assets, In re Celsius Network LLC et al., Case No. 22-10964 (MG), Bankr. S.D.N.Y., ECF Doc. No. 1822, January 4, 2023.

[23] CoinDesk, "BlockFi Files for Bankruptcy as FTX Contagion Spreads," November 28, 2022. https://www.coindesk.com/policy/2022/11/28/ftx-fallout-continues-as-crypto-lender-blockfi-declares-bankruptcy

[24] Securities and Exchange Commission, In the Matter of BlockFi Lending LLC, Admin. Proc. File No. 3-20700, Securities Act Release No. 33-11029, February 14, 2022; Simpson Thacher summary. https://www.stblaw.com/about-us/publications/view/2022/04/06/cryptocurrency-lending-firm-blockfi-announces-$100-million-settlement-with-sec-and-state-regulators-over-unregistered-sale-of-blockfi-interest-accounts-(registered-funds-regulatory-update)

[25] PR Newswire, "BlackRock USD Institutional Digital Liquidity Fund (BUIDL), Tokenized By Securitize, Surpasses $1B in AUM" (BUIDL structural disclosure covering BNY Mellon custody, Securitize transfer agency, and PwC audit). https://www.prnewswire.com/news-releases/blackrock-usd-institutional-digital-liquidity-fund-buidl-tokenized-by-securitize-surpasses-1b-in-aum-302401480.html

[26] Bank for International Settlements, BIS Bulletin No. 115, Aquilina, Lewrick, Ravenna et al., 2025. https://www.bis.org/publ/bisbull115.pdf

[27] Bank for International Settlements / IOSCO, Principles for Financial Market Infrastructures, April 2012. https://www.bis.org/cpmi/info\_pfmi.htm

[28] Tarun Chitra, Autodeleveraging: Impossibilities and Optimization, arXiv:2512.01112, v1 submitted November 30, 2025; v3 revised February 16, 2026. https://arxiv.org/abs/2512.01112

[29] International Business Times / Cointelegraph, "Decentralized Exchange GMX Suffers $565K Price Manipulation Exploit On AVAX/USD Trading Pair," September 18, 2022. https://www.ibtimes.com/decentralized-exchange-gmx-suffers-565k-price-manipulation-exploit-avax-usd-trading-pair-3614630

[30] Blockworks, "dYdX V3 hit by 'targeted attack,' linked to YFI price manipulation," November 2023. https://blockworks.co/news/dydx-v3-targeted-attack-yfi-price

[31] The Block, "dYdX's insurance fund lost $9 million as a result of 'targeted attack': CEO," November 2023. https://www.theblock.co/post/263632/dydxs-insurance-fund-lost-9-million-as-a-result-of-targeted-attack-ceo

[32] dYdX Trading Inc., Post Mortem on SUSHI and YFI Incident, dYdX Blog, November 2023. https://dydx.exchange/blog/sushi-yfi-incident

[33] London Metal Exchange, Independent Review of Events in the Nickel Market in March 2022 — Final Report, January 2023. https://www.lme.com/-/media/Files/Trading/New-initiatives/Nickel-independent-review/Independent-Review-of-Events-in-the-Nickel-Market-in-March-2022---Final-Report.pdf

[34] S&P Global, "FCA, Bank of England launch probe into LME nickel market," April 4, 2022. https://www.spglobal.com/energy/en/news-research/latest-news/metals/040422-fca-bank-of-england-launch-probe-into-lme-nickel-market

[35] London Metal Exchange, "Independent nickel market review" (Oliver Wyman appointment and Final Report). https://www.lme.com/Trading/Initiatives/Nickel-market-independent-review

[36] Bank of England, "Bank of England announces supervisory action on LME Clear," March 3, 2023. https://www.bankofengland.co.uk/news/2023/march/boe-announces-supervisory-action-on-lme-clear

[37] Financial Conduct Authority, "Joint statement from UK Financial Regulation Authorities on London Metal Exchange and LME Clear," March 3, 2023. https://www.fca.org.uk/news/statements/uk-financial-regulation-authorities-london-metal-exchange-lme-clear

[38] Office of Financial Research, Central Clearing and Trade Cancellation: The Case of LME Nickel Contracts on March 8, 2022, Working Paper No. 24-09, 2024. https://www.financialresearch.gov/working-papers/files/OFRwp-24-09\_central-clearing-and-trade-cancellation.pdf

[39] MarketScreener, "Elliott Associates sues London Metal Exchange for cancelled trades," 2022 (litigation initiation); London Metal Exchange, "LME Nickel litigation," case summary. https://www.lme.com/en/lme-nickel-litigation

[40] Stephenson Harwood, "Court of Appeal finds for London Metal Exchange in Elliott / nickel trade cancellation litigation," October 2024. https://www.stephensonharwood.com/insights/court-of-appeal-finds-for-london-metal-exchange-in-elliott-nickel-trade-cancellation-litigation/

[41] U.S. Energy Information Administration, "Low liquidity and limited available storage pushed WTI crude oil futures prices below zero," 2020. https://www.eia.gov/todayinenergy/detail.php?id=43495

[42] CME Group, Testing Opportunities in CME's "New Release" Environment for Negative Prices and Strikes for Certain NYMEX Energy Contracts, Clearing Advisory Chadv20-160, April 8, 2020. https://www.cmegroup.com/notices/clearing/2020/04/Chadv20-160.html

[43] Commodity Futures Trading Commission, "CFTC Staff Publishes Interim Report on NYMEX WTI Crude Contract Trading on and around April 20, 2020," Press Release No. 8315-20, November 2020. https://www.cftc.gov/PressRoom/PressReleases/8315-20

[44] FinanceFeeds, "Interactive Brokers Reports $103m In Expenses Due To Compensations To Customers For Oil Trading," 2020. https://financefeeds.com/interactive-brokers-reports-103m-expenses-due-compensations-customers-oil-trading/

[45] Commodity Futures Trading Commission, "CFTC Orders Interactive Brokers LLC to Pay a $1.75 Million Penalty for Supervision Failures," Press Release No. 8432-21, August 2021. https://www.cftc.gov/PressRoom/PressReleases/8432-21

[46] (Historical baseline note) Public reporting of USDT primary-market redemption pressure during the LUNA/UST collapse in May–June 2022 — Tether's official announcements regarding continued redemption fulfilment during that period, and media records (e.g., CoinDesk's May 12, 2022 reporting). Cited here as a historical anchor for the watchlist; not developed further.

[47] (Historical baseline note) Public reporting of USDC primary-market redemption pressure during the SVB event in March 2023 — official announcements from Circle Internet Financial and Coinbase. Cited here as a historical anchor for the watchlist; not developed further.