Given that it frequently takes a Windows shop most of or more than a decade
to migrate from one version of Windows to another, the duration of this project really isn't too surprising.
As a long-time Linux user and advocate (since 1997, Unix for another decade prior to that), I've long pushed for desktop use, but it's a hard sell at many shops: they've got something that (mostly) works, external partners (businesses, customers, vendors, suppliers) are using (mostly) compatible systems, there's a workforce that's (mostly) familiar with the tools, there's a stable of application-specific software that is (mostly) aimed at that one platform, etc., etc.
Change is hard, especially when the near-term benefits of that change aren't large or obvious.
The Linux landscape has also been somewhat unstable. Not so much if you live on the commandline and don't follow the whims and foibles of leading distros, but desktops, office suites, package management systems, and even preferred distributions have been in constant flux over the past decade. Sometimes there's a lot of benefit is choosing one solution and sticking to it. It's also a fair argument that while the free software licensing and, open source development models have done an excellent job at delivering highly usable standards, tools, utilities, and infrastructures, their success in the end-user, GUI, desktop, and application space has been less exemplary. I still prefer many CLI tools to the GUI alternatives, and basics (text editors) to full office suites. Even now, I only fire up StarOffice, sorry, OpenOffice, sorry, NeoOffice, sorry, LibreOffice, when I need to access a (usually Microsoft-formatted) document, often to print it to PDF and read that instead (using editors to read static content is stupid).
What's making such transitions far more viable is that the desktop monopoly is finally fracturing after over 20 years. I walk into offices today to find what's usually a mix of Windows and Mac systems, with a few Linux boxes for developers. Increasingly you'll find iPads or Android tablets, particularly in meetings. Chromebooks will undoubtedly start to show up. Apple cracked the OS, Google are tackling the office apps space with Google Docs (though as with LibreOffice, I generally use these only once in a blue moon). The SAAS model has gotten people used to using services offered online rather than hosted on their own infrastructure, though post-Snowden security concerns, NSA or otherwise, are definitely affecting this logic. I'm starting my next 20 years of advocacy based on the concept of small, cheap, automatically-configuring service platforms which can provide a distributed SAAS experience without relying on a central provider, think FreedomBox.
Another factor is those legacy industry or nich-specific programs. Increasingly those are not clustered on a single platform, but function only on some legacy (and often unsupported) version of Windows, which must then be kept running, but also somehow secured. Virtualization is increasingly used for this, and with it the realization that no one OS offers access to all the tools a modern information worker uses. I routinely keep multiple VMs at hand to run Windows (often for software to manage my Linux servers), and find that teams will adopt Mac-specific applications that aren't available on either my Linux laptop or Windows VMs. Welcome to the future.
Incidentally, the issues noted in the article with the tight coupling of WINE to Windows applications is an instance of this.
And the truth is that Munich's transition was successful and had strong positives from the start (I've been following it for much of its 13 year life). It was a major transition accomplished gradually with feedback and lessons learned from early deployments applied and adapted as the project progressed.