We determined that the problem stemmed from our use of a Cloud API endpoint that produces a long-running task: a bulk update of issue properties. A logic error in our handling of such long-running tasks caused our HTTP request pipeline to stall. We corrected the logic error and asked customers to restart their migrations.

We conducted a Post Implementation Review and found the response and communications were appropriate. We have updated the process to ensure this portmortem is created in a timely manner.