Every now and then, spurious retransmission requests are received:

NAKACK.handleXmitReq(): (requester=192.168.5.2:4397, local_addr=192.168.5.2:4393) message 192.168.5.2:4393::600 not found in retransmission table of 192.168.5.2:4393: [650 : 1050 (1100) (size=x, missing=0, highest stability=650)]

It turns out that message #600 was actually retransmitted correctly, but the retransmission timer was cancelled too late, so that we got 1 or 2 spurious retransmit requests for the same message.

This doesn't happen very frequently, and only under heavy load (e.g. 8 nodes, every node sends 5M 5K messages with 500 threads).

Note that this issue DOES NOT LEAD TO INCORRECT BEHAVIOR !