EpochAccountsHash doesn't play nicely when the full snapshot interval is greater than the slots per epoch

[brooksprumo]

Problem

I spun up a local test cluster intending to investigate another issue, and then tried creating a snapshot with ledger-tool for one of the nodes to then load on a second node, but ledger-tool would just hang. Investigating the logs points to waiting for the epoch accounts hash calculation to complete before the snapshot can be created.

The slots-per-epoch was set to 8,192, and the full snapshot interval was set to 50,000. I think there may be an issue here based on how the EAH requests and the Snapshot requests are handled by AccountsBackgroundService and AccountsHashVerifier.

Proposed Solution

Debug and fix.

[brooksprumo]

Turns out the reason for this hang was due to AHV not being spun up in ledger-tool, so all ledger-tool commands that replays slots that cross an EAH boundary will hang.