analytics did not recover after out of disk space condition

When running through a sample executing a count(*) while there was a shadow dataset populating, the disk was inadvertently filled. The entire node went offline.

After adding more disk and restarting the node, it appears most services recovered and even the analytics service recovered, but the shadow dataset is not in a good state:
"msg": "ASX0032: Cannot execute request, cluster is UNUSABLE"

The logs seem to indicate no recovery activity. Trying to run this about 45 minutes later still returns the same failure.