Description
The capella fork introduced the ExecutionPayloadHeader field in the LightClientHeader: https://github.com/ethereum/consensus-specs/blob/dev/specs/capella/light-client/sync-protocol.md#modified-lightclientheader.
This allows us now to track the head of the chain without following block gossip or requesting the latest block through request/response.
The node could wait with requesting the block until it really needs it for some call. This should lower bandwidth by a nice factor.
This issue already mentions to potentiall move from follow block gossip to req/resp: #1266
Several models can be used here, and part of the issue is to figure out which ones are still useful for a possible separate modus to run the node in. E.g. Is it useful to follow the gossip in any scenario? That is, some calls (getBlockBy...) could ask for any block, but even then a node might not have it and a request needs to be still made.
Measuring bandwidth usage would also be good to have some ballpark numbers.