Implement support for sending RPC to known but disconnected peer about anemo HOT 3 CLOSED

I think it's ok to make this non-default and require users to opt in on a specific call by, for example, retrieving a different kind of Peer than normal (a RetryingPeer or something).

from anemo.

One more thought, while one could imagine implementing this at the application level by adding retries, that is worse than the network stack doing this because the application doesn't necessarily have visibility into the state of underlying connections. So maybe I'm retrying every 1s but the peer connects after 100ms, and then I'm just sitting around waiting for 900ms until the next retry. Whereas if this is queued in the network stack then as soon as the connection happens, the request can immediately send.

This is indeed a deficiency of the current P2pNetwork::send implementation in narwhal today.

from anemo.

Closed by MystenLabs/sui#5277

from anemo.