I wouldn't need an AB test to know that version A is a better choice, just by the fact that A version is simpler and more objective, it makes me think less than version B before I can take any action.
But hey, some people just need to see metrics to believe.
Well, yeah. How else do you validate your opinion?
That's an interesting point. If you work in a small startup where time/money is a limiting factor, you simply don't have the option to run A/B tests to validate your opinion. In those circumstances, you need to hire good designers who will make good design decisions (which is also not cheap). In lieu of resources, someone needs to make a 'call' on these things.
There generally is a "best" way of doing things but a lot of the time, constraints make it impossible to achieve.
I think A.N. has a point where a good designer will 'know' which design is better out of 2 possible options.
Most of the time quick tests like this are run to disprove the stakeholders involved. In Netflix's situation they have revamped the test again which means the stakeholders cannot/do not want to give into the simplicity mentality.
Just by looking and interacting with it, getting designer peers feedback (design review), performing heuristic analysis etc.
But sure, if you got the time and money, I'm sure Netflix has a lot of these, why not run an AB test–I'm not against AB tests or any other, any information is always valid even if it doesn't aggregate in the end. But in my opinion, A version is clearly a better version than B without having to run an AB test.