During an unrelated conversation today, Aaron and I walked into this code and were wondering whether the overhead/workaround is still needed. This is a placeholder to look into that at some point.
- Memcached alledgedly retaining values too long. I don't know if we've since confirmed this. It seems like the current workaround should be "temporary" not just because it adds overhead, but it means that there are tons of other areas where the same bug is still lurking.
- A desire from Daniel to explore "leaky buckets" (or "moving windows") vs fixed spans for the rate limit throttle.
- Experimental patch: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/624080/1
- Improvements in MemcachedBagOStuff around incrWithInit and changeTTL, and upcoming deprecation of incr() and decr().