Bug: empty protocols after dialing a peer

[juligasa]

Problem

n.p2p.Peerstore().GetProtocols(pid) sometimes is empty right after peer connection. We timeout (20s) before any protocol is returned back.
Turns out that the only two peers constantly failing to show any protocol ID at handshake are:
12D3KooWApxNt9QDA66MqpZyUvZhgkFZUvaNMmcVgJsjeBymhcGC->gabo.es
12D3KooWPubQEomQP49U5x2gynsXKCKrhCoJFjFVmYmNNTbc5nVC->hyperdocs.dev
Restarting hyperdocs.dev has solved the issue for that peer, It turned from failing always to constantly list protocols
The thing is that once a peer does it wrong, it always fails. For example RN I cannot connect to gabo.es in any way

Steps to Reproduce

Expected Behavior

Debug Info

[juligasa]

Now after restarting the local node to switch accounts, gabo.es is working fine and the two new failing peers are:
12D3KooWHAdkjHvQr7UGXFixqJSayYEctsC2JzbYgjzZsWdZnkZk ->kenobi.info
12D3KooWQJfbty2DccQdmkri8CycXA3ZnKuFowWP2n8NiS6QVeF7 -> chelas.es
Seems like every time we reset the node, different failing nodes are created, only resolved by restarting either the local node or the remote one

[burdiyan]

Cross-referencing issue with the libp2p repo: libp2p/go-libp2p#2643