One argument: For many kinds of knowledge work, unnecessary unplanned distractions are a huge drain on productivity. Tapping someone on the shoulder is far more damaging than sending them a message they can reply to asynchronously, even if you're defining expected response time in minutes rather than hours. If there's something that really does require face-to-face time, great, plan a meeting.
Another argument: face-to-face communication is so much higher bandwidth than voice or text chat that it's worth prioritizing. Tapping someone on the shoulder and hashing something out over two minutes can save an hour of online back-and-forth.
Both of these are totally true! It's really about how an individual, and a team, work together best. To me, the frustration is that most modern team environments tend to implicitly choose the latter value system without it being a conscious choice.