Keep in mind that they're really measuring pageviews to Stackoverflow, not actual data about movement. My intuition is that it's probably right, but this is suggestive at best.
Slightly maybe but split tunnel would avoid that. If it was full I'd expect to see an increase from popular cloud hosting regions (if it was a prod vpn) or if corporate VPN's their home offices (which would show as a net zero change)
This data could also signal that people in less traditional areas for software development now have access to remote jobs and are starting to learn more about programming (thus visiting Stackoverflow) to try to get those jobs. The fall in traffic in the traditional places is a signal about movement, but I wonder how much the upticks in other places are a combination of multiple things.
This is the correct (and massive) caveat for everything StackOverflow posts from their own data. It tells you a lot about some developers and absolutely nothing about others.
I use it a lot for my secondary languages, but when writing a well-documented language/framework like C# .NET, I never use it. The Microsoft docs cover everything.
Also, a decent amount of SO traffic is stolen by SO-copy SEO spam sites, too. I used to frequent it, but haven't needed to in years (it seems to more rarely show up in searches for the problems I have, even though I avoid the spam sites, too).