I like to think of test coverage as a negative indicator, i.e. if coverage falls below some defined percentage, it's a bad sign. But it doesn't make sense to optimize for that metric, because, as you said, you can have 100% test coverage but every single test can be bad.

I've always wanted to spend some more time on mutation testing, which can be used to improve test quality instead of just focussing on quantity. But I found it to be completely irrelevant in the industry so far.