Fix Flakey Tests on CI for PubSub examples

[maschad]

Severity:

Low

Description:

On occasion the test-example (pubsub) fails with messages not being received I suspect we could try increasing the delay

Steps to reproduce the error:

Not necessarily easily reproducible as it doesn't happen frequently

[tabcat]

I had similar issues and fixed by listening for subscription-change events to make sure all nodes were peered: hldb/welo@3f60808

for some reason, even with long timeouts, some nodes would never peer

[achingbrain]

Here's another test failure - https://github.com/libp2p/js-libp2p/actions/runs/3891376024/jobs/6641677031

[achingbrain]

I think instead of increasing the delay a better fix in the message filtering test might be for node1 to wait until it sees node2 and node3 in it's peer list for the 'fruit' topic before publishing the messages?

[maschad]

I was originally confused by the logs saying Error: Not enough banana messages - received 1, expected 2 even though two banana messages had been received.

I think instead of increasing the delay a better fix in the message filtering test might be for node1 to wait until it sees node2 and node3 in it's peer list for the 'fruit' topic before publishing the messages?

What's interesting is that this approach doesn't seem to work unless node3 dials node1 directly, i.e. node3 doesn't show up in node1 's subscribers unless node3 dials node1 directly (see 7a6bff3#diff-9483cec2def56f6a8975c97c997d70741698e22fb8a971785d83502f7169bbf1R42)