When tests run in our CI using `flutter drive`, if there is a failure it will loop and try again.

https://github.com/flutter/flutter/blob/434b81f1a58b8528299a63ae7c3ded2aa519e3c9/packages/flutter_tools/lib/src/drive/drive_service.dart#L177-L186

However, it's using the same `device` instance for each iteration. So what was happening was when it would fail to launch, it would tell its listeners that it was cancelled.
https://github.com/flutter/flutter/blob/434b81f1a58b8528299a63ae7c3ded2aa519e3c9/packages/flutter_tools/lib/src/ios/ios_deploy.dart#L486-L489 

Then when the next iteration started, the `vmServiceDiscovery` would immediately return with null because the `deviceLogReader` would be cached from the previous iteration and would already be cancelled. Therefore, bypassing and cancelling the timer.

https://github.com/flutter/flutter/blob/434b81f1a58b8528299a63ae7c3ded2aa519e3c9/packages/flutter_tools/lib/src/ios/devices.dart#L585-L591

https://github.com/flutter/flutter/blob/434b81f1a58b8528299a63ae7c3ded2aa519e3c9/packages/flutter_tools/lib/src/ios/devices.dart#L627

In addition, it seems like sometimes the stop would fail and therefore the the drain would never get the signal that it was done and therefore would hang forever. There was no indication that the stop had failed though because the logs were going to the stream that had no listeners since `deviceLogReader` was already cancelled.
https://github.com/flutter/flutter/blob/434b81f1a58b8528299a63ae7c3ded2aa519e3c9/packages/flutter_tools/lib/src/ios/ios_deploy.dart#L563-L576

Fixes https://github.com/flutter/flutter/issues/127141