Yes, we need to fix this. When we added the push code the fetcher was separated out and created different instance. We could have the resolver cache the base instance created to be used when the fetcher is created, it is kind of a weird pattern though. Alternatively could create a token cache shared almost everything created through the resolver.

These two processes need to remain separate. Ideally, they could use the same token, but it is extremely important that resolution and fetching of content remain decoupled.

Are we going to resolve this or is this a wont-fix? @dmcgowan