Hmmmm... Had a demo from MS about the serverless stuff last week,it's pretty sexy but I was seriously unimpressed with the results, their function created data that wasn't even included in the input parameters - I think you'd have to build a big framework around it to ensure data integrity.
My own view (which could well be wrong and makes me feel old) is that if you want data integrity you're better off directly querying an RDBMS for the foreseeable future, people keep trying to sell me on these magical systems that it's impossible to test because latency. They're great, but as far as they could assure me you're really just pinging stuff into the ether with little in-built reassurance that it'll work. MS will give you a reassurance on data consistency after it's stored but as far as whether it'll actually collect the data in the first place you're on your own.
The trouble, much like the problem with web APIs, is that programmers think this stuff is bloody great, but very few of them can explain how we're supposed to test it, or why we should be shooting everything from a DB, through the application layer, to the api, to the receiving api, through the application layer again and merged/versioned/UID tracked back to the receiving DB when there's a perfectly good standard fine honed over 30 years which gives you a timed snapshot which enforces uniqueness at a point in time and can be optimised against the DB.
But maybe they've never had to reconcile back to the management accounts....
I think I'll just stick an SQL instance on a desktop for now