MYour solution sound like the correct one for this situation.
MWith caching you could reduce the amount of calls, obviously this could result in a situation where the user has an active session in some services for your specified caching time.
MAnd yes, these requests against the introspection endpoint would count against your quota.
MWe get this question quite frequently, both in terms of solution and commercial impact. I cannot promise you anything atm, but we take the feedback seriously for consideration of current pricing model.