I think it's fairly evident that the testing that was made prior to SR1 was insufficient and I can understand why. Projects have limited resources nowadays and unfortunately rigorous testing is one of the first things that gets a cutback. With lowered quality as a natural consequence. The way I see it, that problem can be attacked in one of two ways:
1. See to that the service releases really get tested rigorously which means that organizations must put up the resources needed to do so. The release train must be subject to integration testing.
2. Let the Eclipse users take the hit (like we already do now) and make sure that any problems that are discovered are remedied more or less immediately (i.e. push a new build of the release train or platform without delay). In essence, this would remove the need for service releases.
What we have now is the worst case of all. Virtually no integration testing and when bugs are found, no immediate new release.
I wasn't referring to feature patches with my remark about p2 (I'm not much fond of them either). I was referring to p2's ability to deliver updates in a safe and controlled manner.