Hi!

Thank you so much for fixing #246 so quickly. Unfortunately, it seems that our peerlog problems are not quite over yet.

As of 0.6.2, for fleet durability:

* When a leader restarts, its old session dies, and another node recovers and seals it. * The follower files are deleted only through `adopt()` and `remove_entries_below()` in `node_log.rs`. * This process only runs when there's a new epoch on that follower. * A dead session never opens a new epoch. * Nothing tells the followers to drop their copies.