I would have thought half the problem here is that OOXML isn't even obeyed completely or correctly by Microsoft. Microsoft has also said that it doesn't intend to keep to the standard if it thinks that it can get a competitive advantage by doing so (i.e. "in order to incorporate the latest features"). So even if you read and write OOXML, it's hardly going to make Microsoft strive for better compatibility. In that sense, I would support writing OOXML because then LibreOffice can say "we support the standard that Microsoft buggered ISO with and still doesn't actually support".
But ultimately I see these political gambles not working very well. Has GNU/Linux been a wonderful success story because it was deliberately incompatible with proprietary picture, video, audio, messaging, network and hardware formats? Or has it been popular because it has been able to be compatible without the associated costs or security problems and still provided people with an operating system and software they can share with others? I'd argue the latter. The Stallmans of this world, determined to live by their ideals and eschew every proprietary format and encumbered media, are few and far between.
I support the principle here, but enforcing that viewpoint via software is exactly what all proprietary vendors when they're (perhaps deliberately) not compatible with FOSS. If we rise above that pettiness and support a feature people are willing to program in and work on, then we avoid a kind of tit-for-tat feature-sniping that I think would only hurt us in the end.