as for impact... i don't know that there are great ways to mitigate currently. we've been kicking around an idea for what we're calling an l3 cache that would be a client-side cache that you could configure with behavior when upstream SpiceDB isn't available if you determine that stale results are less of a problem than downtime