I do actually agree with you, what I am saying is that a big part of Linux testing is crowdsourced: you and I and hundreds of millions are testing if there's a regression every time we run Linux, or running buggy, incorrect software on machines in random levels of stability and soundness.

Few unit tests in the kernel would be able to compete with so many chaos monkeys (a reference to Netflix's https://github.com/Netflix/chaosmonkey).