I noticed that deploying (ie promoting) new backend Python3 GAE instances (B1, B2,...) triggers a shutdown of currently serving instances, even while they are still servicing requests.
Is there anything to do about this to avoid getting running requests be terminated prematurely?