The payload cache has an implicit TTL from our action/thread partitions

[jkeifer]

The current data model of payload cache entries depending on the existing and state of action/thread records implicitly sets a TTL on the validity of cache entries where old actions/threads are dropped after some lifetime.

It seems we need to have a table analogous to the thread table as a rollup for events to current action state for the payload cache entries, as a way to roll up actions/action state into the cache entry independent of the actual action/thread records underlying the cache.