No, the proposed change doesn’t seem to be a need. The alternative is for the Timer API docs to indicate that callback functions might be called with the scheduler disabled. If we give the timer implementation that freedom, and publish it in the API docs, then we don’t need to change the code.
The most compelling reason to make the code change is probably to help applications that are currently “getting away with” blocking in a timer callback. Those applications are probably passing all of their tests and seem to function properly, but really there’s a subtle bug that would only show up at tick-count rollover, because the Timer task unexpectedly changes its behavior at tick-count rollover.
Since these applications are already doing something “wrong”, maybe we don’t have any obligation to try to help them out. And since that’s the most compelling reason I could think of, maybe a small documentation update would be better.