Adding an inode limit would limit the number of files/directories that can exist in the tmpfs we use for file IO. Given we already have file size limits, a maximum number of files we'll read, and a timeout, it shouldn't be too important, but it may be worth putting a limit to be safe. Currently it seems to be limited by the cgroup memory limit using (allowing for about 50_000 files with the current config).

This can be done by passing the nr_inodes argument here: