Useful commit messages are valuable, but so is the ability to actually incrementally change and back up your code. As an aside, that's why I think blocking pushes to remote due to failing tests/linter issues/basically any reason is misguided.
However, I’ll often defer a commit, to avoid “breaking the build.”
I’ve come to avoid branching, if at all possible, using tags, instead.
That works for me, as I generally work alone, but it does not work for a team. In that case, we need working branches, and a stable mainline.
In my experience, commit comments are almost worthless, except as “mnemonics.” I tend to use diff views, instead, to figure out commits.
Backing up the code is why I will end up with multiple WIP commits when working on a large change. I want to get it pushed to the remote to not risk losing large amounts of work if something happens to my local copy. It's also nice knowing that if an emergency or something keeps me away from work for some time, the work I was doing could be picked up by a coworker without needing to convince InfoSec to provide them access to poke around the drive of my PC.
Write good commit messages. Use present tense like "Add cats to news feed.". For your "WIP", squash them together once there's something cohesive.
Next on my gripe: A single half sentence email responses!
We used to use Perforce, which forced you to write a commit message. One of my employees used to write "asdf". It got a bit old. I didn't care that much, myself, as he was responsible for maintaining his own code, and he did fine, there.
The Japanese hated it, though, and it probably played a big part in his getting laid off. I learned a big lesson, there, and insisted on my employees being more circumspect, after that.