This incident has been resolved.

The database replicas are having trouble staying in sync. We're moving more subgraphs to the new secondary infrastructure to give the primary infrastructure some breathing room.