During a routine database tuning operation at 8:02am Pacific Time, subgraph failure emails were sent out in error, and users looking at the dashboard might have seen an error status on the subgraph.
The subgraphs marked with errors were in fact healthy, and the error status would have disappeared after 5 minutes.
Subgraph indexing and querying was not degraded during that time.
Our error notification system will be tuned further to prevent this from happening again, sorry for the inconvenience!
