Bitcoin's Rarest Network Event: What the March Reorg Reveals

When Bitcoin's Consensus Engine Runs in Public: The March 23 Reorg

Once in a great while, Bitcoin's underlying architecture stops being theoretical and becomes viscerally visible. On March 23, the network delivered precisely such a moment — a two-block reorganization, one of the rarest protocol-level events in the Bitcoin calendar. For the overwhelming majority of users, nothing felt different. For anyone watching the chain at block height 941,880, however, the Bitcoin consensus mechanism was playing out in real time, in all its probabilistic, competitive, occasionally brutal glory.

This was not a hack, not an exploit, and emphatically not a sign that Bitcoin is broken. It was the protocol working exactly as designed — and that distinction matters enormously for how we interpret what happened.

The Facts

The sequence of events began at block 941,880 on the afternoon of March 23. Multiple mining pools discovered valid blocks almost simultaneously, splitting the network into two competing chain tips ^[1][2]. What followed was a textbook, if unusually deep, reorganization.

At block height 941,881, Foundry USA and AntPool each produced their own valid version of the block. AntPool's version arrived roughly twelve seconds ahead of Foundry's across the network ^[2]. Seconds later, ViaBTC mined block 941,882 building on AntPool's chain, while Foundry simultaneously produced its own competing version of 941,882, extending its alternative chain ^[2]. For a brief window, two fully valid versions of reality existed on the Bitcoin network — two separate chains, each with legitimate blocks, each competing for supremacy.

Bitcoin's consensus rules are unambiguous in such situations: the chain representing the greatest cumulative proof-of-work wins ^[2]. Foundry USA, currently commanding approximately 37% of global hashrate — around 335 EH/s — proceeded to mine six additional consecutive blocks on its chain, through block 941,885, giving it seven consecutive blocks in total ^[1][2]. That dominant run of work settled the matter. The AntPool and ViaBTC blocks were discarded as stale, their miners forfeited their block rewards, and the transactions those blocks contained were pushed back into the mempool awaiting re-confirmation ^[2].

Analysts at b10c characterized the event as rare but entirely protocol-compliant ^[1]. According to data from Hashrate Index, network improvements in latency and block propagation have dramatically reduced such incidents over the years — from approximately 60 stale blocks in 2017 down to just one or two annually since 2020 ^[2]. Two-block reorgs specifically are even scarcer: the last publicly documented instances occurred in May 2023, November 2020, June 2015, and October 2013 ^[2]. A notable separate incident in January 2025 involved three pools simultaneously finding blocks at height 877,991, but that resolved as a single-block reorg ^[2].

The episode did not pass without scrutiny of Foundry's behavior. Mining expert and FutureBit CEO John Stefanopoulos and independent researcher @zawy12 suggested the timing of Foundry's block publications could hint at selfish mining — a strategy where a pool withholds discovered blocks temporarily to gain competitive advantage ^[2]. Specifically, observers noted that Foundry appeared to release its second block simultaneously with its third, rather than immediately upon discovery ^[2]. However, the sources are clear that timestamps alone provide no conclusive proof of coordinated manipulation. With the economic incentives of Bitcoin mining being what they are, overtly malicious reorganization strategies would risk triggering a rapid miner exodus from the offending pool ^[2].

Analysis & Context

To understand why this event matters beyond its technical novelty, it helps to situate it within Bitcoin's broader mining landscape. Bitcoin's difficulty adjustment and proof-of-work design were built to tolerate exactly this kind of fork — the chain always self-heals, and it does so without human intervention. Satoshi's original design anticipated that near-simultaneous block discovery would happen occasionally, and the longest-chain rule was the elegant, game-theoretic solution. What we witnessed on March 23 was not a vulnerability being exploited; it was the immune system responding correctly.

That said, the concentration of hashrate visible in this episode is worth examining seriously. Foundry USA's 37% share is substantial, and the fact that three pools — Foundry, AntPool, and ViaBTC — collectively control over 65% of global hashrate means that theoretical coordination risks, while currently unlikely and economically disincentivized, are not purely academic ^[2]. Bitcoin's security model assumes that no single entity commands a majority of hashrate with malicious intent. It does not assume that a third of the network can never behave opportunistically. Foundry's ability to win a seven-block streak and resolve a two-block reorg in its favor is not alarming on its own — but it is a vivid illustration of the leverage that hashrate concentration creates.

For users and businesses accepting Bitcoin payments, this event is also a timely reminder about confirmation depth. Transactions sitting in those stale AntPool and ViaBTC blocks briefly appeared confirmed before being evicted back to the mempool ^[2]. The industry standard of six confirmations before treating a transaction as final exists precisely because of scenarios like this one. A single confirmation, as this reorg demonstrates, can be misleading. Merchants accepting high-value Bitcoin payments without adequate confirmation thresholds carry measurable risk — not from attackers, but from the natural, probabilistic mechanics of distributed consensus.

Finally, for miners themselves, the March 23 reorg landed during an already difficult period. Bitcoin production costs are estimated at around $88,000 per coin against a spot price near $71,000, leaving many operations underwater ^[1]. AntPool and ViaBTC not only lost this particular block race — they absorbed those losses against a backdrop of compressed margins. It is no coincidence that publicly traded miners like Marathon Digital and Cipher Mining are diversifying into AI and high-performance computing data centers ^[1], seeking revenue streams insulated from Bitcoin's price and difficulty cycles.

Key Takeaways

• The reorg was protocol functioning as designed: A two-block reorganization is exceptionally rare — the last occurred in May 2023 — but it is not a flaw. Bitcoin's longest-chain consensus rule resolved the conflict automatically and without any human intervention ^[2].
• Hashrate concentration amplified the outcome: Foundry USA's ~37% share of global hashrate was the decisive factor in winning the seven-block streak. While no manipulation has been proven, the episode underscores why decentralized hashrate distribution remains a meaningful long-term priority for Bitcoin's security model ^[2].
• Six confirmations is not a suggestion — it's a standard: Transactions in the discarded AntPool and ViaBTC blocks temporarily looked confirmed before being returned to the mempool. Anyone accepting Bitcoin for significant value should treat fewer than six confirmations as provisional ^[2].
• Selfish mining allegations require evidence, not just timing: Circumstantial timestamp data raised eyebrows, but economic incentives and network monitoring make sustained, covert selfish mining strategies extremely difficult to execute without detection ^[2].
• Miners are under structural pressure: With production costs exceeding spot price, the financial stakes around every block win are magnified. Diversification into AI infrastructure signals that pure-play Bitcoin mining economics are under serious strain, independent of any single reorg event ^[1].