After the switch to time-windowed compaction in T133395: Convert RESTBase Cassandra cluster to TimeWindowCompactionStrategy, a rather hacky script was put in place for the largest keyspaces, on a trial basis, to bound the number of SSTables and establish more sustainable tombstone GC. This script continues to run from my (@Eevans) crontab on the cluster nodes.

Though the conversion to time-window compaction has long ago been completed, the issue remained open pending the outcome of this trial, and the "operationalizing" of this script if we were to continue to using it. Given the efforts underway to re-engineer storage, it may not be worth spending any more effort on this, but this ticket can at least serve to document the existence of the hack, and allow T133395 to be closed.