One, two or ten projects cannot possibly reflect people and decision making processes of an entire culture -- The best it can tell you is about that team inside the organization. You could only tell the culture of the company in aggregate, and companies where the F/OSS output is anywhere near the total of what they produce is pretty low.

Would it be completely unthinkable that a company runs an open source showcase team like Oracle runs their America's Cup boat, as an independent entity paid to project the brands name into the world? That would be far from the idea of business open source, but it might not be the worst way of funding the kind of free software that is not directly connected to a specific business case. Many of us don't exactly remember the name Transmeta for their CPUs, for example.

You kinda sorta described SensioLabs: the good team is working on Symfony proper. While the other people are on the web agency side of things and quality is usually not the same.

I can't entirely agree with that. Aside from the possibility of deliberate deception, as discussed by others here, there is always a danger with operating in the open that potentially interested people are only going to look at the latest code, without knowing the history and thought behind it. Many times in my career, I've seen "bad" things done for very carefully considered reasons by diligent and smart people, perhaps to work around identified compiler/OS/browser bugs, or to allow for sensible assumptions that unfortunately turn out not to be correct on some particular platform. The last thing an organisation like that needs is oh-so-smart developers missing the significance of a small comment at the head of the relevant code and judging based on the mustn't-ever-do-this bad practice that was, in fact, based on an entirely rational engineering decision.