you can certainly micro-optimize today if you are aware of the distribution of your dataset. say you know that 90% of resource grants are

reader

. In that case I'd do:

permission read = reader + write

because SpiceDB will evaluate the first element, and then the second one concurrently