so that way

minimize_latency

works is: SpiceDB will choose a revision from the database within a quantization window (default: 5s + some additional leeway) and then use that revision for the request. so long as the caches are at least as new as that revision, they are used