We've deployed some initial workarounds but will continue monitoring and investigating.

We’re observing elevated latency and a higher empty response rate for a subset of queries. We’re actively investigating it.