I resolved my terrible query drama by disabling the health check that runs the query. We have redundant health checks on that data anyway. Now everything is happy again. But I still don’t know why that particular query is forever doomed now.