2-Block Reorg at Height 941,881
March 23, 2026 — a rare-ish fork resolved by Nakamoto consensus
941,881
BLOCK HEIGHT
2
BLOCKS DEEP
7
FOUNDRY STREAK
WHAT HAPPENED
At ~15:49 UTC, Foundry USA and AntPool found competing valid blocks within ~12 seconds of each other at height 941,881. The tie persisted into the next block — ViaBTC extended AntPool's chain while Foundry extended its own. Foundry's chain won, making the AntPool + ViaBTC blocks stale. Flagged by Bitcoin network researcher b10c.
CANONICAL CHAIN
941,879 ──→ 941,881 ──→ 941,882 ──→ 941,883 ──→ ...
Foundry Foundry Foundry Foundry
↘
941,881' ──→ 941,882' (stale)
AntPool ViaBTC
Why — Not an Attack
NATURAL CONSENSUS BEHAVIOR
This was not malicious. Two pools solved the proof-of-work puzzle nearly simultaneously. Network propagation latency meant some miners heard Foundry's block first and others heard AntPool's. The fork persisted one extra block because neither side had converged yet. Nakamoto consensus resolved it exactly as designed — the chain with more cumulative proof-of-work won.
CONTRIBUTING FACTORS
~12 second gap: Tight enough that the second block propagated before most miners switched
~8% difficulty drop: Recent retarget may have contributed to tighter block timing
Mining concentration: Foundry holds >30% of hashrate, increasing streak probability
~8% difficulty drop: Recent retarget may have contributed to tighter block timing
Mining concentration: Foundry holds >30% of hashrate, increasing streak probability
HOW RARE?
1-block reorgs: Roughly every couple of weeks — routine and expected
2-block reorgs: Extremely rare on mainnet — public datasets show only a small handful of confirmed examples
3+ block reorgs: Essentially never observed in modern Bitcoin
2-block reorgs: Extremely rare on mainnet — public datasets show only a small handful of confirmed examples
3+ block reorgs: Essentially never observed in modern Bitcoin
What It Means
MINING CENTRALIZATION
Foundry USA mined 7 consecutive blocks (941,879 through 941,885). Foundry and AntPool together control nearly half of global hashrate. While this reorg was benign, the concentration increases the statistical probability of extended block streaks and competing forks. The event reignited discussion about pool dominance and Stratum V2 adoption for decentralizing block template construction.
CONFIRMATION SECURITY
No double-spend occurred, but any transaction with only 1 confirmation could have been temporarily displaced. This is a practical reminder that low-confirmation acceptance carries real risk. The standard 6-confirmation guideline for high-value transactions remains sound. Merchants and exchanges accepting 1-conf should take note — especially for Lightning channel opens.
THE PROTOCOL WORKED
This is textbook Nakamoto consensus. The network saw two competing chains, miners built on whichever they heard first, and the chain with the most accumulated proof-of-work won within minutes. The protocol did exactly what it was designed to do.
Selfish Mining?
stratum.work saw no Foundry jobs for their winning 941,882 — what gives?
THE RED FLAG
b10c noted that stratum.work (which monitors pool job distribution) did not see any stratum jobs from Foundry mining on their winning block at height 941,882. This raised the question: was Foundry selfish mining — withholding blocks and mining privately to gain an advantage?
b10c'S CONCLUSION: PROBABLY NOT
The fee rewards in the stale blocks were tiny (0.008 + 0.017 BTC) — not worth the reputational risk. b10c demonstrated with a Bitcoin Core functional test that normal network relay with geographic distribution can look like coordination. Foundry was likely mining "in one region" while stratum.work was connected to another, explaining the missing job data. Murch: "The natural explanation is plausible… the community tends toward conspiracy theories perhaps a little too enthusiastically."
STREAK FREQUENCY ANOMALY
Analysis by Sorukumar on BNOC showed 7+ block streaks are 3–7× more frequent than hashrate alone predicts, with Foundry at 3.17× expected frequency. This suggests pools have structural propagation advantages — they see their own blocks first and start mining immediately. Not necessarily malicious, but a real centralisation concern.
HISTORICAL 2-BLOCK REORGS
From b10c's
stale-blocks dataset, confirmed 2-block reorg examples on mainnet include heights 180,804 (2012), 261,198 (2013), 358,998 (2015), 656,477 (2020), 788,686 (2023), and now 941,881 (2026). The last two gaps were ~132k and ~153k blocks — roughly 2.5–3 years apart.