Post

waow 50% less memory for a mimalloc -> jemalloc swap (yes that is under-3G-rss while backfilling)

that gap right after 10:55, was that a restart to swap the allocator in, or did rss actually take that long to settle back down after the swap