Got a question about caching and potential hotspots. I'm writing a schema that allows for custom role definitions, and it looks something like this:

definition authz/principal_types {
  relation includes: orguser:* | identity/group:* | resourcemanager/organization:*
  permission supported = includes
}

definition authz/role {
  relation has_resourcemanager_organizations_get: authz/principal_types
  relation has_resourcemanager_organizations_update: authz/principal_types
  ...

For every supported principal type, I'm writing a relationships like

authz/principal_types:1#includes@orguser:*

. I'm using this to avoid having to add a wildcard for every single

has_*

permission relation in my schema, and to make it possible to add a new principal type by simply adding a new

includes

relation and writing a new relationship. The assumption is that the includes relationships will be cached so lookup will be fast, but one concern that's been brought up is that this might result in a dispatch hotspot due to how sub problems are cached. Is this concern accurate, and if so how concerned do I need to be? Is there any sort of adaptive spread for frequently accessed sub-problems?