TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Even an on-chain rollback can't recover stolen assets?

Even an on-chain rollback can't recover stolen assets?

2026.09.02
Share
Three Public Chain Emergency Shutdowns: Revealing the Practical Limitations of On-Chain Emergency Powers.

By Liam 'Akiba' Wright

Translated by: Saoirse, Foresight News

Within four days, three blockchain networks sequentially halted block production. Each network shutdown invoked entirely different emergency powers, with only Cronos rewriting part of its official chain history.

According to Cronos, after the Tectonic protocol was exploited via a vulnerability, validators shut down the network through consensus, rolling back the chain to its state prior to the attack and resuming block production from block height 90,896,189. This operation not only halted block production but also directly altered the chain state. All transactions and state changes that occurred after the recovery point were no longer part of the restarted mainnet.

Ontology and ICON adopted a different set of contingency measures. Ontology suspended block production before confirming malicious activity; its update note on September 1 stated that the malicious activity did not result in user asset losses. ICON first paused the exploited contract before shutting down the entire network; the foundation claimed that during the migration phase, it controlled the network, at which time most of the stolen ICX had already been transferred to exchange custody accounts.

Blockchain shutdowns are merely the first layer of control. The deeper questions are: Who has the authority to order a network shutdown? Can they rewrite confirmed chain states? Once funds cross chains or enter centralized custodial institutions, which losses become irreversible?

Network-triggered event contingency measures | Disclosed authority info | Known recovery risks | Cronos Tectonic vulnerability exploit | Halted network and reverted to pre-attack chain state | Validator consensus; restart announcement omitted vote counts and thresholds | All on-chain activity post-checkpoint invalidated; funds sent to Ethereum fall outside Cronos control; final loss accounting for Tectonic protocol pending | Ontology Routine inspection flagged potential risks, later confirmed malicious activity | Preventively suspended block production, no rollbacks conducted | Core developers, tech team, and validators involved; emergency trigger threshold undisclosed | Transactions unexecutable during patching/upgrades; no user asset damage detected | ICON Replay vulnerability in migration contract | Paused contract first, then shut down entire network | Foundation controlled network during migration phase, core validator count reduced | Losses covered by foundation; recovery of ICX held in exchanges hinges on custodians, legal procedures, and law enforcement

Comparison of emergency response measures across the Cronos, Ontology, and ICON public chains

Cronos: From Shutdown to Rewriting Chain State

Cronos termed the incident response an "validator consensus emergency action." The August 31 restart announcement indicated that at 23:49:01 UTC on August 30, the network resumed block production from block height 90,896,189, rolling back the chain state to before the Tectonic vulnerability attack.

Cronos's shutdown operation implies making allocation decisions regarding the recovery point. After the checkpoint, vulnerability-related on-chain states, along with all unrelated transactions during that period, were wiped from the mainnet. The restart announcement came without transaction lists, validator statistics, voting weight thresholds, or participant node rosters. Cronos promised a post-mortem report that must fully explain the response process and technical impact scope.

Even the scale of assets actually protected by this intervention remains undecided. TRM Labs estimated that after TONIC token prices were manipulated, approximately $75 million in assets were borrowed out; about $6 million flowed to Ethereum, and roughly $68.7 million were rolled back within the Cronos chain. Bitquery's figures show a higher total outflow, with approximately $8.3 million moving to Ethereum and a total of 10,961 blocks being discarded.

The two statistical methodologies track different subjects, and Tectonic's official final loss figure remains pending publication. However, one thing is clear: Cronos's rollback can only restore states remaining within its own chain; assets on the Ethereum chain are completely beyond its control.

Tectonic's asset resolution plan still leaves user account issues unresolved. The protocol stated it would prioritize opening withdrawal and loan repayment functions while halting deposits and new borrowing. This approach provides users with exit and deleveraging paths, but whether liquidity providers can fully redeem their funds remains unconfirmed. The pending post-mortem report from Tectonic must still clarify the vulnerability mechanism, total capital outflow, bad debt scale, recovered assets, and remaining liabilities.

Recovery progress across various infrastructure layers is not synchronized with the chain consensus restart. Cronos warned that protocols, cross-chain bridges, block explorers, and RPC services require more time to recover. Alchemy's status page also separately logged this shutdown and subsequent restoration. While the chain network may declare an official restart, dependent services may not yet be ready.

Ontology: Shutdown Only Buys Time for Resolution, Does Not Revert Transactions

Ontology's response action occurred before confirming malicious activity. The network stated that the core development team discovered potential security vulnerabilities during routine inspections, immediately suspending block production to allow the tech team and validators to conduct a system review.

A September 1 update announced that the review confirmed malicious attack activity; the mainnet will remain shut down for vulnerability patching and network upgrades, noting that the attack did not compromise user assets. Ontology's goal is to resume normal operations within 24 hours, contingent on successfully completing security scans, patches, upgrades, and testing.

Ontology's shutdown preserved all confirmed on-chain states, only halting the confirmation and settlement of new transactions. The announcement specified no recovery point nor disclosed any sets of transactions to be invalidated.

Publicly disclosed authority information is incomplete. The announcement mentioned involvement from the core dev team, tech team, and network validators, but failed to identify who holds the final binding decision-making power or provide numeric emergency trigger thresholds. Ontology's VBFT documentation outlines its standard consensus mechanisms, including nodes generating confirmed blocks and managing contract updates to the consensus node set, but the document only covers normal operational scenarios; the emergency pause rules used on August 31 were not made public.

Even without causing asset loss, the shutdown still incurs real costs. Ontology informed users that on-chain transactions would be unprocessed and advised against executing time-sensitive operations; it later noted that network restart depends on patches, upgrades, and testing. Users could not adjust positions or settle transfers on-chain, forcing all external services integrated with the chain to wait for network signals.

The criteria for restoring operations are safety-oriented, but specific details are limited. Ontology stated it aims to resume service within 24 hours once fixes, upgrades, testing, and verification are complete, though it did not disclose who determines condition fulfillment or what the exact trigger thresholds are.

This creates governance-level uncertainty: while the announcement names participating parties, the entity with final authority to decide on restart remains undefined. For users, the current risk stems from service interruption rather than confirmed asset loss or chain rollbacks.

ICON: Why Blockchain Shutdowns Came Too Late

The ICON incident fully illustrated the entire process from alert generation and response to assets escaping chain-level control.

According to the foundation's post-incident review, attackers replayed two historically valid signed withdrawal messages 1,492 times between 02:01:02 and 02:21:12 UTC on August 27. Precision flaws allowed 1,490 of these calls to succeed, draining 119,866,000 ICX and 531,600 bnUSD from the foundation's treasury pool.

Monitoring systems triggered alerts at 02:08, but technical investigation began afterward; the affected contracts were paused at 03:53. Major exchanges gradually halted ICX deposits and withdrawals at 05:54, with a full network shutdown taking effect at 06:18:54. ICON completed its restart around 07:51 on August 28, approximately 25 hours later, simultaneously patching the underlying vulnerability.

The review attributed the root cause to incident response workflows rather than inadequate detection capabilities. Alerts triggered within seven minutes, but similar warnings were frequently confused with unrelated RPC anomalies, so the system failed to notify on-duty personnel. Technical investigation did not start until around 03:40, shortly after which contracts were finally paused.

By the time the chain officially shut down, most of the affected ICX had already been incorporated into the exchanges' own custody systems. ICON's on-chain control measures could not prevent exchanges from transferring or converting the assets they held. The foundation could only rely on exchange asset freezes, preservation notices, lawyers, and law enforcement agencies for remediation.

Custody boundaries directly dictate loss attribution. ICON stated that all affected assets belonged to the foundation, with ordinary users' deposits, balances, and holdings untouched. The report shows 531,600 bnUSD and 1.366 million SODA have been fully recovered; out of 113,634 lent USDC, 82,430 were reclaimed. Confirmed net losses amount to approximately 150.2 ETH, plus 31,204 USDC. The vast majority of involved ICX were merely frozen or tracked by exchanges, rather than genuinely recovered.

ICON's governance architecture also differed from the other two cases. The review noted that the foundation controlled the network during the token migration; migration guidelines stated consensus operated in maintenance mode with only 7 core nodes. Therefore, this shutdown relied on a distinct operational structure explicitly governed by the foundation.

Emergency Powers Are Essentially Balance Sheet Authority

Every blockchain shutdown essentially shifts risk to different entities.

  • Cronos modifying official chain history: Can protect assets still within chain jurisdiction, but invalidates normal on-chain activity unrelated to the exploit, leaving it powerless over assets on Ethereum.
  • Ontology converting risk into time costs and service availability losses, leaving transactions unsettled during investigation without confirming any book value asset losses.
  • ICON completing contract and network isolation only after assets had already moved out of chain custody; confirmed losses are borne by the foundation, with recovery of frozen ICX reliant on exchanges and judicial authorities.

A mere decentralization score masks these vastly different outcomes. A more pragmatic evaluation metric asks: Are emergency response rules publicly disclosed? What are the trigger thresholds? Does it merely halt new blocks or rewrite confirmed chain states? When intervention occurs, who controls assets that have escaped the chain's jurisdiction? Who commits to covering residual losses?

Full post-mortem reports remain pending from both Cronos and Tectonic. Ontology needs to disclose attack details and emergency authorization rules, followed by confirmation that upgrade and restart criteria are met. What truly deserves comparison are the risk boundaries drawn by each network — which histories, timelines, and capital pools are placed at risk.