What is really interesting to me is how are they doing persistence.
Node and SQLServer? Sounds very harsh.
MongoDB? does that run on Azure?
My bet - their Node servers proxy to the old servers via dedicated API for data. Node should be fine at doing these requests due to its async nature.
Even when you run with Express, to build such a feat requires a TON of infrastructure (did you notice the X-Tracking, X-Response, etc headers?).
Much of the open source/public infrastructure in the Node ecosystem feels raw (example: loggers. did you see how many loggers there are? do you think they're production worthy?).
Though I would accept this and fix whatever I can (took me a while to find a proper hierarchical logger that is also async, for example), I find it hard to grasp how Microsoft would.
This makes me assume they built everything from scratch.
Didn't say they're far from each other, of course it may be possible. I'm also aware of Microsoft's support for Node and Azure, and their contributions to Node.
Last time I wanted to integrate postgres with Node, I had flashbacks of myself, 10 years ago, building an ORM before ORMs were that hip - this is true for any of the impedance mismatch offered by taking an RDBMS and plugging it with Node. I think there's something with using stores such as MongoDB with Node that feels very right, almost essential.