AntoMoreau and JoshC189 We continue to work with relevant teams to figure out how, what and if there are additional changes that need to happen in this area. I understand that there is a bit of ambiguity, but I do not have anything more to share at this time.
We'd be very interested in understanding if you are seeing perf issues after disabling token cache. As every environment will be different, there is a possibility of this, especially if the environment was already running a bit "hot" before the change. We are not aware of significant impact of this change, though. It would be helpful if a ticket was opened if this change made a significant perf difference so we could investigate and have a look at perf logs before and after the change. Performance issues are especially tricky as variance in client use / mix between environments can make all the difference.