That's not what it is about though. There is plenty of evidence that there are pros and cons both to WFH and work-at-the-office, assuming the work lends itself to work-from-home to begin with. This is at best a datapoint and not so much a grand conclusion worthy one at that.

The metric 'code productivity' alone is such a terrible one. I remember the 80's when such things were introduced. The best one that I ran into professionally was 'object code size' (because we don't want to count those pesky comment lines as production now, do we?). It didn't take long for the rookie in the team to outscore everybody else based on those metrics. He found the largest library in the system to link to...

In general I'm against such metrification of productivity and in software I'm more against it than in other industries because I think software quality is a very hard thing to measure to begin with. Lines-of-code and such are useful on an assets list during a business transaction in a descriptive way. But they're not very useful in other contexts.

As for the code review data they analyzed:

"We find that sitting near coworkers increases the online feed- back that engineers receive on their computer code. Engineers ask more follow- up questions online when sitting together, and so, proximity can not only increase in-person but also digital communication. Proximity is particularly integral to the online feedback received by young and less tenured engineers. "

I've seen the exact opposite happen as well. Proximity decreased the feedback because there was no need to communicate formally what could be communicated informally.