Drag and drop layouts are alive and well(ish)
This is very common in the enterprise space. Especially in the realm of operational dashboards and business intel apps.
As with many things in the dark world of enterprise software, the user experience is rarely ideal. In the case of a complex and broadly applied product like Tableau, users are willing to take on the learning curve to get more from their data. There aren't many of those users so Tableau has to structure their pricing accordingly: the drag and drop creator is just an admin serving a larger group (and the per user cost is high).
It's not for everything
There are two big costs to configurable layouts: engineering effort and user learnability.
Obviously, there will be more development and testing time invested in giving users the tools to configure their own interface and to make that interface reliable (no body wants to work on that QA team!).
From the user's perspective, you're asking them to learn how to get at what they want. In a world of hyper-focused, "dive in and start winning" products, configurable is more hype than feature.
The web portals that Steve Jones mentioned in the comments are a great example of a place where it seemed right but just didn't stick. Those products were pretty awesome. I created some nice start pages over the years and lamented their demise. But the reality is that most people just couldn't be bothered.
IME, even with enterprise products, you often have a set of norms rather than an infinitely variable mix of needs. If you do provide a configurable UI, you still need to do the work to find a good set of default "starting points". The defaults or "canned views" represent aggregations of your users with like concerns or workflows (personas, if we have to label it).
In practice, these defaults end up accounting for 80–90% of the required applications. Now you have to ask yourself, is it worth the effort to build out UI customization for a minority of your users ... and risk complicating things for everyone else? The answer to that question is usually "no".
As with most scaleable products, the average often kills off the outliers.