Hmmm... There is no more exception here, but it still works wrong... Now it silently drops active subscriptions.

The original pull request: #2222
The original issue: #2221

@kantora I can't merge this PR, because we don't have any test for "HandlePrune" neither in MNTK tests nor in unit tests. Could you create the one? If not, just share some examples of code, which can help me to reproduce this bug

@alexvaluyskiy As for me, it reproduces just each time cluster pub/sub is used, if you give it at least 2 or 3 minutes to run. And, to be honest, this pr just hides the problem (as it prevents the exception to appear) and don't fixes it. I've described the main issue in #2226.

As for this particular PR - the fix is obvious from code as the previous version modified dictionary while was iterating it and fixes the recently merged PR #2222.

@kantora I've tested your code, it really fixed the problem. But we shouldn't test our code manually. So we should create a test for it. I've created an issue for this case
#2232