I would be fascinated to see your evidence that in-house code is any better on average than open-source code.

I haven't done a lot of consulting lately, so I haven't seen much in-house code in the last few years. But my experience is that the average in-house codebase is worse. And that makes sense from the incentives. Open-source projects that want more than one contributor need to be approachable enough that people join in. Whereas with most in-house code, people commit to working on it without ever seeing it. Switching to work on another open-source project is easy; switching to another job is hard. Open-source authors get to decide when to release; in-house code is generally driven by execs. And so on.