Stalled BEEFY report

[svyatonik]

After adding BEEFY to Rialto test deployments recently, we've observed BEEFY stall (just after deployments were restarted), with some interesting logs. I can easily repeat the same logs locally (by running scripts from https://github.com/svyatonik/parity-bridges-common.test), but has never seen yet stalled BEEFY there. Logs are:

...
gossip: Registering peer who=<PEER1> protocol=/paritytech/grandpa/1 role=Authority
gossip: Registering peer who=<PEER1> protocol=/paritytech/beefy/1 role=Authority
...
gossip: Registering peer who=<PEER2> protocol=/paritytech/grandpa/1 role=Authority
...
beefy: for block 9 public key missing in validator set: XXX
...
beefy: Missing validator id - can't vote for: YYY

So it seems that

1. nodes are sometimes fail to open "/paritytech/beefy/1" connection (see above - PEER1 has registered two protocols and PEER2 has only registered one). To me it seems like some race at networking level - need to explore lower-level logs;
2. <missing> things are strange, but they are DEBUG logs, so probably not an error.

[acatangiu]

I believe this is fixed in paritytech/substrate#10882