The articles base premise is tenuous at best. It assumes that “unknown” has a best case attribution to Linux with no backing for that assertion.

Unknown could just as easily be windows, chromeOS or macOS or just automation for that matter. Why would only Linux report as unknown for only a portion of users?

Given that the Unknown line is directly mirroring the ChromeOS line, it’s much more likely that it’s misattribution from ChromeOS. (And yes ChromeOS is Linux under the hood but the distinction matters because of the implication of the article)

What is the methodology of the statistics collection? Is it just user agent strings?