Wasted data in range_blocks_and_blobs_requests

[dapplion]

Description

Recently added metrics on sync show a decent count of unknown sync requests for range_blocks. This line here

lighthouse/beacon_node/network/src/sync/network_context.rs

Line 354 in d6ba8c3

metrics::inc_counter_vec(&metrics::SYNC_UNKNOWN_NETWORK_REQUESTS, &["range_blocks"]);

For example, during sync in sari it happens between 0.5-1/min

I believe this is due to the design of range_blocks_and_blobs_requests, which will drop its item if any of the two internal requests fail. Hitting the unknown request condition causes fetches data to be wasted.

Allowing each internal request in range_blocks_and_blobs_requests to be retried would reduce that waste rate. See

• Improve range sync with PeerDAS #6258