Random number generation, and especially decentralised such, is not something I’ve looked into much.
Server side code running on vaults to deliver something, is a whole problem of it’s own; decentralised applications. The latter, yes we can definitely build and run such even today. The vault software is a decentralised software whose job it is to store data. This already runs.
If the latter will be a solution to the former, that I am currently not informed enough to make any good comments on, but it would surely present options to do so.
As a business idea you mean? (The OP)
There is already a key-value database functionality inherent. When it comes to the type that many would first think of - relational db - I think the network currently lacks some solutions to be able to implement it. (I started a topic on it not long ago here. The EventStore that I have been building on is very rudimentary, and the performance is not the best in the current design, but sure enough works. But I actually think the best event stores, in non decentralised environments, are built on top of SQL databases. Why? Because SQL databases are very hardened and have extremely powerful features that make them so much more useful than a simple key value store, or append log, for real world use of event streams.
Databases with a more advanced functionality require computation, and so to make it truly decentralised you have to solve that part. I write about it in the other topic.
As for business idea, well sure. There will be someone with enough team, vision and capital of various sorts, to provide more advanced database solutions if the tech allows it. And looking at what has been built so far the tech for it would be there eventually. There are still some limitations, but those tend to go away with time.