bug: complete channel wind-down on failed channel in retry loop

[AndyTWF]

At the moment, if we attempt to do a channel wind-down and it enters the failed state, we don't wind down the other channels, we need to do this to conform to lifecycle atomicity requirements.

┆Issue is synchronized with this Jira Task by Unito

[sacOO7]

This is exactly same as #405 right 🤔
More like addressing 2nd case as per #405 (comment)

[AndyTWF]

Yes - just splitting it out so we can track it separately :)

[sacOO7]

Found one more case where we don't do channel wind down in the first place : (
Mentioned at the comment -> #405 (comment)

[AndyTWF]

Found one more case where we don't do channel wind down in the first place : ( Mentioned at the comment -> #405 (comment)

This instance is ok - all the other channels are already wound down above, the only one that won't be is the one that just went into "failed".

[sacOO7]

I didn't get what you meant, I don't think we are doing explicit wind down for other channels for mentioned case

[AndyTWF]

I didn't get what you meant, I don't think we are doing explicit wind down for other channels for mentioned case

Line 406 does a wind-down of all other channels, then the promise waits for the one-remaining one to re-attach. So if that remaining channel becomes failed, all then other channels are already sorted.

[sacOO7]

Ah, got it. Seems we try to attach only when channel wind down for other channels are completed.
Btw, Line 406 does partial wind down as per second case mentioned at comment, so we need to fix that one too