We were on the M2 shared tier and experienced 2+ hours of downtime with no access to our cluster/replicas, even while the status showed that Atlas clusters were not affected... just the "operations on the clusters." Anyways, time to upgrade to an M10 dedicated instances, I guess.