In my case, the runs are often run in parallel and in batches of varying sizes, so "same name" or rotation schedules wouldn't work well. They are a kind of hybr

In my case, the runs are often run in parallel and in batches of varying sizes, so "same name" or rotation schedules wouldn't work well. They are a kind of hybrid log/storage file (they contain intermediate data sectioned/logged into the actions that created them, like an extremely verbose log). Dedicated SSDs would be very nice, but out of the price budget (i.e. its a zero cost project, run on existing scavanged hardware!). As you say, /tmp is working well for me, my example was to perhaps illustrate that there are (what I think are) legitimate uses for large files in /tmp.