Doh! I didn't log my times and had forgotten about that that table. Thanks for the reminder.
Though I don't have the earlier execution times to verify I can see we did have a few unloads today of some of the table columns potentially used by this query on some hosts.
I've run a few more times now, will my colleagues mostly logged off, and the times have been consistent. I'll keep my eyes on it and let you know if I ever see it happen when there are no applicable entries in M_CS_UNLOADS.