Possible resources leak with non-closed p2p connections

[algorandskiy]

Received a private report regarding p2p nodes:

Doing a goal node status showed it was still online and synced with the network, but then checking htop showed a whole bunch of algod threads had been spawned and sleeping. I'm guessing something specific to P2P is causing a goroutine to get stuck or not exit properly.
It doesn't seem to be P2P-only nodes, but any node that has P2P and/or Hybrid enabled.

Trying to be clever I thought maybe I could see which location was referenced the most:
grep "\.go:" systemd_algod_output.log | sort | uniq -c | sort -nr | head -n 50

  22 github.com/algorand/go-algorand/network/wsPeer.go:517 +0x394
  22 github.com/algorand/go-algorand/network/wsPeer.go:516 +0x354

systemd_algod_output.log

Nov 29th:

Since updating to 3.27.0 on all my nodes and monitoring the thread count, they don't seem to be inflating like before. Had them running for at least 2 days now and they're at anywhere from 17 to 25. Before they were going up to like 150-220 threads.
They are all in hybrid mode, and one is p2p-only.

Dec 18:

I've only just gone back to check the thread count since 3.27.0 and it seems it wasn't resolved after all.