Cannot restart a beacon node after a few days using the `--enable-state-diff` flag.

[nalepae]

After running a beacon node for a few days with the --enable-state-diff flag, the beacon node cannot restart.

Step to reproduce the issue, with Prysm V7.1.6.

1. Start the beacon node from a fresh DB with the --enable-state-diff flag
2. Wait for a few days
3. Restart the node (gracefully)

After the restart, the

[2026-07-10 11:35:28.22] DEBUG node: Starting State Gen

log is displayed during ~10 min, then:

[2026-07-10 11:44:23.11] FATAL main: unable to start beacon node: could not start modules: could not start state generation: load state by root: could not get ancestor state: slot 13636387 not in db due to checkpoint sync: cannot retrieve data for slot

Then, every eventual restart shows:

[2026-07-10 11:44:27.24] ERROR node: State-diff database corrupted; restart aborted. Delete database and re-sync from genesis/checkpoint. error=state-diff database corrupted

Full logs

Additional note:

• I use the pruning flag, and the backfill flag. This may probably play a big role here, when we look at the old age of the 13636387 slot.
• This is the second time this issue happens on my node.