fix(ngc): dont double every message, if we are not directly connected

[Green-Sky]

but we and the other peer would support direct.

---

All hail the network profiler.

(missing returning udp traffic)

---

This change is 

[JFreegman]

I definitely think there needs to be more leeway than this. UDP connections can be weird and take a bit of time to be established. What's the actual source of this bug, anyhow? Alice knows Bob's IP, Bob knows Alice's IP, but for whatever reason they can't establish a UDP connection?

Edit: And no I don't think it affects hole punching.

[Green-Sky]

I definitely think there needs to be more leeway than this. UDP connections can be weird and take a bit of time to be established. What's the actual source of this bug, anyhow? Alice knows Bob's IP, Bob knows Alice's IP, but for whatever reason they can't establish a UDP connection?

Yes.

Edit: And no I don't think it affects hole punching.

Good to know.

[Green-Sky]

Change of plans. I took a look and NGC simply has no NAT hole punching code, making this double sending act as such.

So we need to replace this with some proper code, either by copying the existing code from dht, or writing something new.

I will lower the timer to once per second and push this as a quick fix to get this into the release.

[JFreegman]

Reviewed 1 of 4 files at r1, 2 of 2 files at r2, 1 of 1 files at r3, all commit messages.
Reviewable status: complete! 1 of 1 approvals obtained

[Green-Sky]

Closing in favor of #2905 , not sure why "checks/check/check-release (pull_request_target)" keeps failing with this one.

(had to recreate the commit, so the cache is not pulled over to the new pr)