Alright, resolved with no downtime.
I didn't realize that the upgrade to Mastodon 2.6.0 *removed* the old rake tasks when tootctl was added. So, we weren't cleaning remote images. This caused the disk to fill up at 2018-12-19 15:10 UTC. I had mistakenly set up an alert for "Utilization" rather than "Disk Capacity", so Server Monitor never warned me about this.
* Add correct alerts (done)
* Update cron job to use tootctl (done)
* Ensure cron failures email me