I recently deployed an experimental proof-of-work protocol called Shardhash. The idea is simple: miners compete by finding hashes with a required suffix pattern, and rewards scale with the number of matching hex digits.

Something unexpected happened during testing.

A wallet that had never interacted with the protocol before submitted a failed claim and then a successful claim. The successful solution was Tier 10 (roughly a 1 in 16^10 event), with a nonce value of 18.

The interesting part was the timing:

- The failed claim used nonce 17 but was rejected because the round had already expired.

- The failed claim was submitted 300 seconds after the round started, even though the submission window is only 12 seconds.

- The next claim used nonce 18 and succeeded.

- The successful claim was submitted 480 seconds after that round started (42 Ethereum blocks later), about 180 seconds after the failed submission.

At first glance this looked strange because a very low nonce satisfying a Tier 10 target is extremely unlikely if only a small sequential search range is tested.

After investigating, the more likely explanations are:

- the miner used a custom implementation rather than the reference miner,

- the miner was able to precompute because the challenge became available before the round opened (extra 12 seconds window before submission period starts),

- or it was simply a rare probabilistic event.

Shardhash contract: https://etherscan.io/address/0x295121422b9d0fd3cbddc9e203ae9b4a1eff0082