If I show each user of a website a highly customized, user- and workflow-specific page for the same url based on context and previous activity, then I have to generate it every time ("hand crafted, artisanal web pages") so the weight of the backend is now proportional to traffic.
Amazon.com tries to split the difference. You and I see the same bones of a page for an Apple Watch 10, but little bits load in and show you laundry detergent and me pickles. But Amazon makes more money every time I click on pickles and add to my cart, so there's an expectation that the fractional penny they pay to load the page fragment results in more sales. That doesn't work the same for SaaS applications, so you need to use even this trick sparingly, not build your whole product value statement around it.
For the control issues crack, the paying customer (not their users) is attracted by all the levers and dials, but cannot appreciate the cost using them exposes them to. The development cost can amortize over time, increasing your profit margins and letting you recoup the R&D costs, but the cost of keeping a cluster running cannot. And you've painted yourself into a requirements corner you can't get out of. Eventually their eye drifts to competitors with fewer high-cost, high-value features in favor of low-cost.
Is Gmail a highly customized website? What about Atlassian suite?
Is gmail viable as a standalone company?
Does showing different snippets of existing texts for each user,and providing sorting and searching functionality is highly customized in a way that will usually be too costly on the backend to make viable money?