Problem description

This is what happened if I don't extract roots for all messages (te/minimal_exchange_gossipsub_message branch)

on a holesky node subscribing to all subnets, this helped reduced max 0.4% of gc time

we extract roots of all possible messages, especially beacon_attestation for:

• if extracted root is known (skipped slot), we don't queue
• if extracted root is not known, queue messages by root then by slot
• actively search unknown gossip block by root

Solution description

considering the gossip validation logic of beacon_block:

• it checks if proposer index is correct
• it checks if proposer signature is correct
• it checks if any blocks are known for the same proposer in the same slot

it's a guarantee we never have 2 blocks for the same slot passing through gossip validation

• still extract roots for some first messages of slot
• once gossip blocks are processed, no need to extract roots anymore
• for skipped slots, gossip block never comes, we'll keep extracting roots like the current flow

Additional context

No response