Two nodes, NodeA and NodeB.

NodeA has flush_throttle_timeout = 100ms
NodeB has flush_throttle_timeout = 10ms

NodeB sets persistent_peers=NodeA

Things work fine, but NodeB repeatedly hits flush failures, causing the connection to break. It happily reconnects shortly after due to persistent peer.

If we change NodeA's setting to match, then the connection stops dropping.

It's not immediately clear why we lowering the timeout on NodeA makes it easier to keep up with a lower timeout on NodeB