Due to provisioning problems, new api instances weren't created properly, leading to an under-capacity from around 16:00. This under-capacity resulted in a backlog of processing transfers.
When we resolved this issue, a 2nd backlog was created for pending emails. The database handling this backlog could not cope, and was replaced by a beefier one, causing a delay in the sending of these email. After the database was replaced the backlog finally resolved itself at 2am on April 7th.