To me it seems like the options for user-defined schema are either push-down the schema (and do one schema per spreadsheet) which lets you take full advantage of your DB's query planner statistics system etc, do an entity-attribute-value table and deal with building your own query planner and collecting your own statistics, or go with a document store and build some kinds of dynamic indexes instead of dynamic table schemas.
That being said, I generally agree: I’ve been bitten by enough database products that are some startup’s main product to really trust them. I’d rather use a database that’s either maintained by something like the Apache Foundation (Druid, Pinot, etc.) or one that’s a side project of some big company (FoundationDB) or one that’s been around long enough to be unlikely to surprise us when the team is acquihired.
I've never seen database that either (a) allows you to create that many schemas at all or (b) doesn't become less stable because of it.
One example of this is that the database may create a directory structure per schema which will then result in file system performance degradation or hitting end user limits.