Hi all! We've had SpiceDB set up in a Kubernetes cluster for a while now, but we're finding that when we are rolling over the deployment, a small number of calls to SpiceDB seem to get

UNAVAILABLE: End of stream or IOException

errors in that time period. We've been careful to configure our client facing load balancer to time its drains and terminations appropriately, so we're curious what this might be caused by... If I had to guess, I would say it might be pods being terminated before completing dispatch requests, but I'm not 100% sure...