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.

We've seen a spike of 503 errors at 2:35pm CET.
We're investigating