Previous incidents

All task inserts and updates were rejected during 11:55:57 to 11:58:02 UTC due to a incorrect migration.

Migration is over -- we have seen very little impact on production.

We're no longer seeing dropped requests.
We identified heavy load of an un-optimized dashboard query. We will migrate this query to a newer, faster query.

We reduced the load by 3x on a problematic client, hoping this would help mitigate the circumstantial issue in the future.

We have identified the root cause and implemented a fix.

We haven't been able to isolate a root cause and it looks like random traffic congestion due to slightly increased load.
We'll continue monitoring such spikes and make scaling adjustments if such issues happen again.

The fix has been deployed. We'll keep monitoring.