The example of Amazon's mandated RTO has been much discussed elsewhere as a notable exception to that company's normally data-driven approach. As I saw commented elsewhere:

"If we have data to show that mandated RTO is more productive then let's share and act on that. If we're going with opinions then let's use mine"

The elephant in the room here is that 'the data' doesn't exist, and for good reason: 'Productivity' is subjective in most jobs to begin with, and even where it isn't Remote vs Office is not a significant variable.

Consider some 'traditional' proxy measures for productivity- punctuality and attendance. It's pretty clear that if someone isn't at work then they aren't productive and many employers will put employees through formal processes to dismissal in response. Why don't we see this being highlighted with remote work? Should be easy and obvious to measure and demonstrate right? Could it be that these things improve with remote work?

Consider also the pandemic impact on remote work and that we are talking about a return. It's clear that many organisations could manage perfectly well, as they themselves have proved.