Pretty sure all these companies are heavily A/B testing every change they make, and would immediately abandon any new design if it decreased engagement or retention even slightly.

These changes are "hated" by a vocal minority who loves to complain about change to show how much the liked the old version. They don't quit using the product, though. And if you ask their opinion again even three months later (which is another thing big companies tend to do), they'll have completely forgotten what the difference was between the two.

Of course, some people do quit a service after any given change, no matter how innocuous—they just disappear without saying anything. But changes, even change for change's sake, also tends to increase signups. So in practice, it's just a question of whether the positives outweigh the negatives of the change. And only real usage data can answer that, not customer surveys—and especially not customer exit surveys.