Wherever we were successful in establishing good policies that were followed, we had a couple of highly invested individuals, and we did code reviews across the shared codebase (vs only restricted to one's areas of ownership).

I think both points are important, but the latter is harder to achieve in practice: people lazy out by "optimising for efficiency" by not writing proper MRs, asking close colleagues for reviews, and thus missing out on broader shared knowledge, and coding standards (I much prefer "style") deteoriate.