We're going to run out of space on /dev/events and /dev/metadata/udev/events if we run for too long, since a process can die without cleaning itself up. We'll need to either use runfs directly, create eventfs from runfs (optimized for deque-like access patterns; see [1]), or create a "reaper" process that runs every X minutes to clean up dead event directories.
[1] http://comments.gmane.org/gmane.linux.devuan.devel/2888