Since I actively supervise my testing, and actively test my tests, I don’t find the prospect of mistakes in my test code to be terrifying.
I don’t understand why the possibility of difficulty with test code should cause me to write shallow, lame test code.
That's why I've made it a habit to write only strictly linear tests in my own code, and use fixtures and parametrization to keep duplication to a minimum.
Ultimately, there's no one true best strategy. Everything is full of trade-offs all the way down.