I’ve put together a catalogue and a corresponding post detailing 143 Bitcoin blocks that met the required proof-of-work target but failed a consensus rule, from the 2010 overflow to two previously unreported F2Pool timestamp failures from 2026!!
The catalogue brings together the surviving evidence for each case, so anyone can independently verify which consensus rule the block broke.
A few findings:
- The two F2Pool headers from April and July this year have timestamps roughly 7 and 14 days below their parents’ median time past. Both were found in merge-mined auxiliary chains; neither has a known Bitcoin-side capture.
- Auxiliary chains preserve 49 of the 143 headers, and provide the catalogue’s only recorded observation for 35. They check their own rules, so they can preserve evidence of candidates that Bitcoin peers reject.
- 89 blocks contain the same P2SH-invalid transaction, mined between April and July 2012. Archived transaction lists and preserved headers establish its inclusion in each.
- 127 of the 143 have no complete body. The surviving headers, coinbases or transaction inclusion proofs, together with chain context, still establish the failures.
- Another 10 reported cases remain separate because the evidence is insufficient.
The same merge mining research also recovered roughly 1,400 stale headers new to the stale-blocks collection. Merge mining has proved to be a productive side channel for both stales and invalids, and monitoring those chains is essential to a holistic Bitcoin monitoring effort.
In my merge mining monitor tool you can step through all invalid-blocks discovered in the merge-mined record (note that invalid blocks are currently called error blocks, this was the language I originally used):
If you have old node files, pool logs or other records that could fill a gap, contributions and corrections are very welcome.
It's a pity that I linearize my backup chain data because that excludes orphans, so I cannot help with this. Could put a bounty on the full block data: maybe someone still has an original older datadir that contains some of these?
I didn't realize there were this many merge mined chains:
I love your monitor, very cool data here.
Yeah, there’s a few things that stand out.
Qbit was launched just a few months ago, has strong backing.
Some of chain in Recovered Datasets are actually still running, they just have so little hash behind them that they contribute very little. At some point I will elevate them to their true status as Live Sources, but it’s not a priority. Terracoin is an example of this, seems to be only mined by Mining Dutch, who land a block every few months.
These chains have downtime, have been hacked etc. Elastos was down for days a few months ago.
Given F2Pool offers FPPS on BTC do they eat that loss?
Yeah, FPPS pays out regardless of what the pool finds right?
TBH I am not even sure how aware F2Pool are of the losses of these blocks.. Would be interested to get comment from them on this matter.