hello. in our loadtests we're observing long durations for latencies for the namespaces queries. this is probably because our schema is very large and most of the relation definitions are concentrated on a small number of resource definitions. we don't understand why it's necessary to load the namespace definitions repeatedly after a node has started, rather than only once on startup. it would be useful to just keep them cached but it seems like the namespace cache is not configurable independently. is there another way to achieve this?

spicedb treats namespaces just like relationship tuples, and they only live for 2x the quantization window. We added a new feature that uses the Watch API to maintain an up to date cache of your namespaces, it should solve your problem.