Well I fear that when you offer a functionality that seems so easy, eg pollMultiple = true (eg just set a boolean to true), then people may fall into a trap, and think it just works as easy as well.
Even if we document that this option has many caveat and you should use it with care, we see many people on the mailing list, that dont read javadoc / eip docs / and whatever. And this is not something we can do much about. Its just how the world works today.
About your use case, to drain a JMS error queue, then the current code is actually worse, as the consumed messages will be acked on the JMS broker one by one, so you can lose messages if Camel crashes, or the likes.
I think we should consider having a new EIP for this use-case as the current pollEnrich is designed to consume a single message, eg to be a counter-part to poll which is actually producing to an endpoint.
Until we have a comprehensive implementation that can support a new EIP / DSL, I suggest to start with a helper class (as you say), and then work our ways there.
The current code is too weak, and open up a can of worms for the end users.