axyl

A failed batch doesn't always need to be executed again. In @Rayls Axyl, the reason for failure matters.
I explored the Axyl source code to understand how transaction nonces, retry limits and restart recovery work together.
1. Retry only when the conditions match
Axyl distinguishes between "NonceTooHigh" and "NonceTooLow". A transaction with a nonce that is too high may be waiting for an earlier transaction, while a nonce that is too low may already be stale.
The orchestrator requests a retry only when batch digest v2 is enabled and every transaction in the batch is counted as "nonce_too_high".
2. Retries have a limit
At the inspected source revision, "MAX_NONCE_TOO_HIGH_RETRIES" is set to "3".
When the retry limit is reached, the registry keeps the batch digest registered instead of removing it for another attempt.
3. Restart recovery protects deduplication
At the same revision, Axyl scans up to 2,000 recent canonical blocks to reconstruct batch digests after a restart.
The recovery logic also checks whether a batch can safely be retried. When a database read fails, the digest remains registered rather than being removed automatically.
My takeaway
Retrying is only one part of reliable execution. The system must also prevent duplicate processing and restore its deduplication state after a restart.
These findings are based on source inspection at commit "23976578fe2ff81924923cf69f8bebb9ce2a45bf". The values describe that revision and may differ from the latest code.
#Rayls #Axyl #blockchain #Web3