Currently, we're using an async CAN_HOP message sent on peer-mux-established to any incoming peer to identify relay nodes. This is inefficient and requires adding a timeout after a connect sequence to give nodes time to mark a peer as a relay. There are several discussions and proposal to mitigate this, but it fits the theme of service discovery well (e.g. libp2p/specs#32).

This issue is here to document the reason for the timeouts in connect sequences and a reminder to change it as soon as service discovery is here.