twoeths

yeah this is a good finding, it may not be an issue for now but could be in the future if someone want to use Buffer instead (to improve performance,...

@g11tech take a look at https://github.com/ethereum/consensus-specs/pull/2954

> @tuyennhv can this PR be changed to 'ready for review'? @wemeetagain this is not ready, we need to take time to test it (nodes are busy with testing mesh...

It shows that Unknown Block Sync now triggers very frequently due to unknown block root gossip attestation, but most of the time gossip blocks still come before Unknown Block Sync,...

> How does this interact with the cache for attestations that reference blocks still not seen? This does not interact with that cache, it just help finding the unknown block...

PR in progress https://github.com/ChainSafe/lodestar/pull/4114

this could be due to the result of #4011 in v0.37.0 - old mcacheLength is 5 and with a heartbeat interval of 1s => we keep message ids in mcache...

From v0.37.0 (18h00), there are more late messages deliveries There are also more broken promises Gossip block process time is higher (this is known after we update to esm module)...

it's worth to also mention that state transition is more unstable in v0.37.0

So I deployed v0.36.0 to contabo-0 and contabo-20 and compare to v0.41.0 (beta group, contabo-12, contabo-13). Below is number of peers with score > 0 (mesh threshold) - v036.0, contabo-0...