How much test coverage is enough?
Coverage is a map of what you executed, not of what you verified. Chasing the percentage produces tests that assert nothing.
A codebase at 90% coverage where every test calls a function and asserts it did not throw is worse than one at 55% with tests that describe behaviour, because the first one gives false confidence.
Coverage tells you which lines ran. It has never once told anybody whether the software works.
Sam Kettleborough, performance engineer
Cover the paths where a bug would be expensive, and let the rest be whatever it is.