Some of the machines in the RESTBase Cassandra cluster have had the reserved space disabled (set to 0) on the volume where Cassandra storage happens, others have the standard 5%, which on a ~5T volume amounts to ~231G. Since Cassandra is the only thing using these volumes, it would seem sensible to remove the reserved space, and to make this a standard (and documented) part of the configuration. I'd be interested to hear if there are any objections from Ops on the former, as well as any suggestions about how best to do the latter (Puppet?).
I don't think there was ever any controversy over whether we should do this. AFAIK, it didn't happen because it wasn't obvious (to me at least) how, and because of a general lack of time (or relative priority, etc).