[patrec]

From their post-mortem: https://blog.bambulab.com/cloud-temporary-outage-investigati...

> The result was that the print job was successfully completed on the printer, but our cloud service believed it had not been done. When service resumed, the once-jammed job was resent, leading to the unexpected printing of an already finished job.

That's why you make message sends in a distributed system idempotent, unless you have a really excellent reason not to. Whatever generated the print job when the user pressed “print” should have generated a unique job id, and when the printer was resent the same job id it should just have replied “Done”.