It's a nice, pithy answer to give, and in general you shouldn't rewrite a backend "just cause". However, Wunderlist's backend was built with the intention of being a bit of a prototype, and the skills on the team are vastly different from when it was written (in PHP) compared to where we are now (using mainly Rails with a little Node.js in the backend).
In addition, contrary to Spolsky's claim, unmaintained code does rust, or "rot". Libraries you were using might go unmaintained in newer versions of a framework that fixes critical vulnerabilities. Code that held up well to thousands of users doesn't necessarily when you're talking about millions of users. Fighting code/app entropy is a real thing.
But we're taking steps to make sure this backend neither suffers from the "second system effect" (we got that out of the way with Wunderkit ;-)), nor simply needs to be rewritten again in a year or so. Spolsky is correct that in general, code is harder to read than write - which is why we're writing Wunderlist 2's backend to be easier to read at the cost of extra work up front. Strict conventions, enforcing documentation, challenging any code that isn't the clearest way to express the problem it's solving - well, we'll have an engineering blog post to discuss our "Pull Request Parties" and other related process improvements in depth at a later point in time/
The short version, though is that the WL2 backend is probably the most beautiful, readable codebase I've ever had the privilege of working with. Of course I'm more than a little biased, but hopefully the proof will be in the pudding when we launch it out into the world.