resources references are local, having performance advantage over jndi remote lookup, but that may not be substantial. More importantly, some app server (e.g., JBoss) does not work with remote lookup. Normally Websphere works both ways in my experience. So treat this as a guideline instead of a hardcore rule.
we use such setup and I find this very useful.
You indeed initially need to update your ear with listeners info which will/might need involvement of external teams.
Yet, at a point of time, your listeners list will be complete and it won't require anymore any external team, except maybe from time to time, a queue update.
Then, with this setup, when we want to isolate our application, we deploy the default plain pega ear, without listener information.
If you also disable agent startup (and file listener if any), you can then bring up your application isolated, not consuming or processing data. When our PEGA work is completed, we just redeploy the ear with listeners and enable back agent.
We just got sudo access for stop/start/deploy and we're almost fully standalone.
Having this technical setup externalized also helps for easy handling of all our environments. We don't need to consider the non-prod and prod setup inside PEGA, it's outside.
So you intially need external team to gain more freedom later (if you can receive few sudo access)