This came to our attention last week and was fixed then. Basically, when we resolved the domain names for API server backends, there was an edge-case that could crop up when a DNS lookup temporarily failed. This led to us temporarily not routing API requests to that specific API backend, which is obviously problematic.
More details and the fix here:
NREL/api-umbrella-gatekeeper@bdb2be2