Catchup regression after p2p merge due to rate limiting

[onetechnical]

Subject of the issue

Requests for blocks in normal catchup are triggering rate limiting. This results in slow catchup performance and repeated pauses.

Note that this behavior starts with the p2p merge c8407abc #5939. A binary built from the immediately preceding commit 63c0d5bd does not have this problem.

Your environment

• Software version: 3.26.2066.nightly [rel/nightly] (commit #9364f13b)
• In this particular case, running on DevNet
• Replicated on both Ubuntu and Darwin

Steps to reproduce

1. Get the latest nightly binary
2. Start the node

Expected behaviour

Fast, progressive catchup.

Actual behaviour

• Ledger will advance a couple hundred blocks
• Ledger will stop advancing for about 18s
• Repeat

This is due to rate limiting, as the logs will report 429 errors, e.g.:

HTTPFetcher.getBlockBytes: response status code 429 from '/v1/devnet-v1.0/block/ilk0f'. Response body '' 

On the server side this will be reported as "Remote IP Connection Rate Limit".