Currently we archive buckets under memory pressure or if we open a new bucket due to time forwards/backwards. We definitely want to continue to archive buckets under memory pressure, but we may wish to close buckets rather than archive them when we have time forwards/backwards.

If we keep these buckets eligible for query-based reopening, then we would retain good bucketing behavior in out-of-order scenarios, but we wouldn't track truly idle buckets as archived unnecessarily until we run into memory pressure.