We've determined that student writing wasn't being autosaved from early Sunday morning through Monday about 3:30 PM EST. Students who submitted signal checks had their work saved, but students who didn't lost work. Initial investigations show that this was due to an automated database upgrade early Sunday morning, but the autosave service didn't adapt correctly. Alerts were raised but not caught and responded to by the engineering team. We'll be looking into fixing the root cause as well as improving our alerting.
We take data loss like this very seriously and recognize that it has a big impact on students. Fixes to the root cause will make sure that we store student work if at all possible so we can at least recover it later.