Agreed with a lot, but for many companies I think cleaning up git history is more effectively accomplished by cleaning up PRs than actually trying to clean up the commits themselves that developers write.

"Good commits" are one of those things that benefit the whole, but no individual developer is incentivized to do. Because of that, it's easier to enforce at code review (rather than hope developers just do it). Many companies do this by squashing changes, so a developers' commits and the commits on trunk are decoupled.

I think the best way to make changes (and therefore commits on trunk) understandable is to split them up into series of dependent changes; something supported by the code review tooling used at places like FB / Uber / Google, but not super well in GitHub... Which leads to large PRs with a lot of unrelated fixes.

Disclaimer: I work at a company trying to bring that dependent change / stacked diff workflow to git and GitHub (https://graphite.dev).