There should be a process in place for routine testing on each target environment. Testing on BeOS shouldn't just be done because Bob happens to like doing his development on BeOS.
I didn't say that was the only testing, either. However, use during development is also testing, and in practice it can easily add up to more hours of testing than routine testing - and if all of those hours are on the same platform, then you get skewed picture overall. In addition, routine testing is usually much more scripted.
But frankly, your original point doesn't make sense to me, because there's simply no downside, so why argue about upsides? The team can agree on a single code style (or have that mandated from above), but how does that affect editors, and especially OSes? The only meaningful constraint is that one's choices must be able to handle the mandated style.
Same thing for other stuff. For example, it makes sense to mandate a testing framework. But if multiple IDEs support that framework, what's the point of forcing the choice on that?
