Given a bounded scheduler policy that drops running
task instances, if there's an immediate return in
via a later perform call, the state update from the earlier
task instance may get applied later than the task instance
with the immediate return.
Applied a similar approach that Glimmer uses, which is an
incrementing tag, by which we will track which state updates
are still valid. Anything after the last applied tag will be valid
and state updates from before will be ignored.
Covered by the included test case based on #422.
Given a bounded scheduler policy that drops running task instances, if there's an immediate return in via a later perform call, the state update from the earlier task instance may get applied later than the task instance with the immediate return.
Applied a similar approach that Glimmer uses, which is an incrementing tag, by which we will track which state updates are still valid. Anything after the last applied tag will be valid and state updates from before will be ignored.