Logs are too noisy

[Sword-Smith]

The current problem is mainly WARN massages from libp2p and its integration. But in general, we'd like:

• exactly one INFO message per new block (new tip)
• exactly one INFO message per new peer connection (upgraded to peer_loop)
• exactly one INFO message per closed peer connection (closing of a peer_loop)

Edit: added clarification

[aszepieniec]

This policy is rather under-specified.

• For each policy, do you mean no more than one INFO message, or at least one info message?
• What about logging on other events? Related to mining / transactions / syncing ...

[Sword-Smith]

This policy is rather under-specified.
...
* What about logging on other events? Related to mining / transactions / syncing ...

You're right. A problem that's even worse than too much logging is when you think the application is frozen because it's not responsive, when in fact it's doing some initialization job at startup.

I like our warnings on all peer punishments. Those are important to track as punishing too much in some cases. If you're composing, guessing, or proof upgrading you certainly want log messages about that on an INFO level. Same goes with transaction initialization.