This morning (between approximately 10:10 and 10:40 UTC), we experienced a performance degradation of email ingress and egress in our Dublin2 and Dublin3 clusters.
This issue was brought to our attention by automated alerting built on our internal observability tooling. Application scaling policies increased capacity during this period, reducing its overall length. This was a repeated issue, seen previously on 16th March, caused by abnormally expensive processing of a set of malformed emails sent to a large number of recipients.
Mitigation steps we have taken so far have unfortunately not been sufficient to avoid a recurrence, and as such, we will continue to investigate to improve resilience.