Triage can be the hardest part of learning to review

Some of the more consistent characteristics of the scientist are the inclinations to teach, fix, trouble-shoot and generally help out their colleagues, students or research trainees. As with any GoodThing, when it is taken to an extreme it can turn into a BadThing.
A little Twitt action around the toobs recently brought this into focus in the context of reviewing manuscripts submitted for publication in a journal.
In this particular case it was a question about the writing quality in the manuscript.

Science peeps: is it sufficient as a reviewer to state that the manuscript (a lit review) is simply not written in readable English?

One response, from the always perspicacious Dr. Isis, supported the helpful side. She suggested that it is not cool just to shrug off the authors with such a complaint and that one should address what one can as a reviewer. That’s my interpretation of the 140-char / time discussion anyway. The other response, from your heroic and very humble narrator, was that it is perfectly acceptable to kick a manuscript back for being written in an unreadable manner.
What I’ve landed on as a discussion point for today, however, is the notion that we would ever seek to simply triage our efforts on a peer-review assignment. Is it permissible, within the polite society of the Tribe of Science, to essentially refuse to review one component of a paper (say the data) until another (essential, optional, peripheral?) component has been fixed?
I say it is. Indeed, I think that one of the primary functions of a mentor trying to help a graduate student or postdoc learn to review a manuscript is giving them the tools for deciding when NOT to review the manuscript in any depth. When to decide that there is a fatal flaw that simply must be fixed before it is worth going into any additional effort.
This could be be because of the data- a missing condition or control or perhaps a completely inappropriate (or missing) statistical analysis. Not worth going into the arguments in the Discussion until you agree the data actually present some sort of knowledge, right?
And yes, we can acknowledge right off the top that the line of when to say when can be a very slippery thing. Hard to nail down any hard and fast rules making this a very subjective aspect of review. Of course, any scientist should feel free to make any comments they like, even to the extent of trying to rewrite the dang paper themselves. I just think it is bad strategy and a waste of time.
So while I understand that any trainees who are starting to review manuscripts* want to look like they are insightful and pay strict attention to detail, I also work on getting them to understand that it is perfectly okay to triage a complete piece of garbage.
__
*You mentors do have something of a plan in place for starting this process with your postdocs, right?