The compiler switch would be for emulating (or for disabling) the global switchover that we want to take place at some point. We currently have the wrong default, so to speak, and we’d want to change that at some point. Then it is useful to already be able to do that in earlier releases, and also to be able to revert to the old default in later releases.

Otherwise, I agree that it should be in the source. In practice, looking for the presence or absence for T? vs. T! will be a good heuristic for whether the code assumes the new or the old semantics.