Investigate more graceful shutdown of the InMemoryChannel

[Cali0707]

Problem
Currently, we are seeing flaky upgrade tests due to the IMC losing events. This is because the IMC is not persistent, so some events are lost in the upgrade process. We should handle this more gracefully by ensuring that all events have been handled before the channel shuts down, and indicating on new incoming events that they were not handled because the channel is shutting down.

Persona:
Which persona is this feature for?

Exit Criteria
The upgrade tests are no longer flaky from lost events

Time Estimate (optional):
How many developer-days do you think this may take to resolve? 5-7 days

Additional context (optional)
Add any other context about the feature request here.

[Cali0707]

/assign

[Cali0707]

cc @pierDipi - did I miss any of the points from the WG discussion on this?

[Cali0707]

@pierDipi do you think we can close this now that #7415 is merged?

[pierDipi]

Yes, thanks for the improvement in #7415