Don't try to reconnect the socket when we closed it intentionally

[nickvergessen]

Extracted from #3292

[nickvergessen]

/backport to stable18

[danxuliu]

I have added some fixes on top of the current strategy.

However, I have realized that the whole beforeunload handling could be removed if we just check the code attribute in the close event. In theory when the code is 1001 the user is navigating away, so if we do not try to reconnect in that case it should be enough.

Right now, if the user tries to navigate away, stays in the call and the external signaling server goes away in that moment then it will not get reconnected. Sure, a very corner case, but possible nevertheless. This would be prevented by checking the code instead of using the current approach (although there could be other drawbacks I am not seeing, I do not know).

[nickvergessen]

Maybe I mixed soemthing up, but the real problem I had which is why I did this was when the connection is terminated because we switch the HPB because of the clustering. But maybe the code works there as well

[danxuliu]

Yes, but the commit is covering two different cases: prevent reconnecting when the signaling is disconnected, and prevent reconnecting when the page is about to be unloaded.

As far as I understand it, prevent reconnecting when the signaling is disconnected is what is needed for the HPB clustering. That is OK, and it is not related to my comment. I only refered to prevent reconnecting when the page is about to be unloaded, as I think that could be handled easier just by checking the code from the close event, with no needs for beforeunload events, timers and so on.

[nickvergessen]

Feel free to add/change it

[nickvergessen]

Updated and much less intrusive now!
Thanks for the pointer to the close code @danxuliu

[nickvergessen]

The commit Fix signaling toasts not being permanent should be dropped in favor of a rebase since #3449

[nickvergessen]

Rebased and dropped it

[danxuliu]

Tested and works 👍