This property is necessary because on native we were clearing threads based on the returned updates. For thick threads, the updates aren't returned, so we didn't clear any threads based on the result. In thin threads, we were checking the returned updates.

Changes to this function are quite complicated, but it should match the previous behavior:

1. We're cleaning threadID before a request is dispatched
2. Then we dispatch a request
3. Then we analyze thin thread result and invilidate more threads

We're unifying the behavior by:

1. Returning empty invalidation array from thick thread protocol and conditionally invalidating based on the length
2. Moving catching errors outside a dispatched action promise - the behavior should be the same

We don't need to disatch because it is included inside the protocol