Open
Description
At the moment when a message times out in the reprocess queue, we try to process it again. For attestations this is almost certainly a waste of time, because if they were processable we would have already dequeued them when their matching head block arrived.
I want to try experimenting with just trashing attestations that timeout, rather than trying to reprocess them. This shouldn't happen a lot during normal operation, but may happen on unhealthy nodes, and during non-finality or in other periods of high load.