As a "scientist" mentioned in your comment (they call us Research Engineers), I have to tell you that this is a business decision. One or more than one suit in the C-level made this decision.

And, as a scientist, you cannot build a model with a never-happened-before phenomena. Only "risk specialists", "risk engineers", etc. which are different names for BSers will even try to "predict" this.

How can one build a "model" with no x (password policing enforcement) and only y (stay or drop) is beyond me. If anyone does, I will show you an idiot/BSer.

I am sure this is a business decision. Or MLEs/Deep Learning Engineers are going beyond the capabilities of their fields.

Sure, you can predict if a user is going to drop if a show he was binging is going away or his watching pattern, time spent watching, etc. But to build models with password policing enforcement? Hah!

Trying to predict whether this decision is harmful in the long run is not in the realm of science, but of art, business, or bullshitting.

I don't know if NFLX did a smaller scale pilot study and tried to build a model, but that would be inadequate.

Note: Never worked for Netflix.