But that's an interesting facet in itself. The more common refrain you hear in corporate settings is "we need a ground-up rewrite" - how did these projects manage not to need that? I have my suspicions, but it's a very interesting question nonetheless.

As a big fan of rewrites (but not as big a fan as some), I think Linux has avoided needing big rewrites with a lot of very clever modularity in its design. Many corporate software programs do not have that level of modularity. Modularity has performance costs, is slower to develop, and "looks ugly" on a block diagram, but it pays dividends for open source projects and projects that need upgrades over time.

Individual parts of Linux go through relatively frequent rewrites (eg filesystems), but not the overall framework.

Linux IS corporate. It has been corporate for a long time - the heaviest commits come from paid, professional developers.

Final decisions are made by totally independent folks like Linus, Greg or Andrew. They set the rules and also the community, corporate has to adapt to them or go away.

Also, what happens is that rewrites happen progressively, in small steps.

Not, I think, in the sense meant here. You're right, of course, but the distinguishing factor is that "corporate software" typically only involves a single corporation. There's no (or little) balancing of interests.