I recently got a manuscript to review from a specialty journal I occasionally publish in. It came at a time that I was very busy and I only briefly looked over the abstract before accepting it. I've had a lull of reviewing that has lasted about 2 months, so I figured it was time to put in my dues and the paper sounded interesting. I wish I had read the abstract more carefully because the text is a train-wreck and I probably would have been able to tell that was going to be the case if I had not been rushing. In this particular case, the data suck almost as much as the grammar and I don't have any problem tanking it. However, grammatically putrid papers that have a story to tell always leave me conflicted.
On the one hand I can only imagine how difficult it is to write coherent science in another language. It's tough enough for natively English speaking scientists to write science well and I'm pretty sure my career would be over if I suddenly had to write everything in Japanese for my science to get noticed. I give scientists from other countries, especially ones where the system of writing is fundamentally different from ours, a lot of credit for even turning out something that is readable, let alone coherent. As a grad student and even early in my post-doc years I often spent hours correcting all of the grammatical issues in these papers, with some reviews reaching in excess of ten pages. If the science was good and the only thing holding the paper back was the communication of the results, I felt like I could be a major help to foreign scientists by spending the time to make comments to improve the way the paper reads.
These days I just don't have that kind of time to devote to being a high school grammar teacher. I will happily critique the science of any paper I am working on, but my patience for dealing with the communication side of things has dwindled dramatically. I have sent several papers back in the last year or so with the critique "this paper can not move forward until it is heavily edited, preferably by a native English speaker". I realize that this is a significant hurdle for foreign authors, but I have never really felt that it is the reviewer's duty to correct grammar and if the shoe were on the other foot, I would ensure that my work in another language had been proofed by someone who speaks the language I am writing in. I have done this for colleagues in the past and will if asked again in the future. But I see that as very different than my editing responsibility as a reviewer.
I'm curious how others deal with the good science but incoherently written papers that come from foreign authors (realizing that the same can happen from English speaking authors, which is a different problem). I'm sure there is a range of reactions.