Thanks for the responses. I actually started to write this at 8 this morning, but quickly got sidetracked.
Ultimately, what happened was that 2 of our 4 hosts disconnected and I was unable to reconnect. I was able to ssh in and run commands, including to restart the management services, but I still couldn't connect. One of them did not have many vms, so we rebooted the host and restarted the VMs. Then we migrated all vms from one of the good hosts and rebooted, then repeated that with the other good host. That left us with one host with several vms, one of which was a windows terminal server running some stat jobs I didn't want to interrupt. I scheduled a reboot time over the weekend for that, so all 4 hosts were good after that.
Here's what I'm still trying to figure out. I've removed datastores in the past two different ways and got essentially the same result. One way is to follow the 2004605 article. The other is to just delete a ds without unmounting first, refresh the storage view to make sure the ds has disappeared, offline the volume on the EQ and rescan storage on all the hosts. I've done both many times, and the only difference I've found is that the first method takes a lot longer. I only do one ds at a time.
My co-worker did two things differently. First, he unmounted but didn't detach before deleting the ds. Second, he deleted a lot at one time from a host before rescanning and getting into the EQ to offline the volume.
The real irony is that my coworker is usually extremely thorough and cautious. I guess everyone has their bad days.