Unnecessary repeated epoch transitions when node is stuck

[dapplion]

Describe the bug

Due to the invalid head state root issue, our Prater node got stuck for a few days in limbo, memory is stable. However, many epoch transitions are done per epoch (32-50). Maybe the regen is not caching the state after advancing it? Or doesn't know how to query it.

Check in the REST API metrics the only external interaction with the node is getProposerDuties, there was no gossip received during that period.

Note: prometheus query for epoch transitions per epoch

increase(lodestar_stfn_epoch_transition_seconds_count[384s])

Expected behavior

Steps to Reproduce

Screenshots

Desktop (please complete the following information):

• OS:
• Version:
• Branch:
• Commit hash:

[dapplion]

Closing issues for specific incidents on old versions of Lodestar, please re-open if it happens with latest stable