DataSet API versus Stream API

Background
I posted a question earlier that had way too many parts, so I'm breaking out individual questions. Hopefully, these are easier to answer now and easier for other people to discover later. I can figure much of this out by testing myself, but am hoping others already know and can answer without much effort.

Question

Our likely cases are modestly sized data sets updated on a regular basis. These are not one-off data uploads, neither are they huge uploads. This leaves me wondering if we should be looking at the Stream API instead of the DataSet API. Stream sounds better for continuous and reliable updates, but also seems intended for high-volume pushes.

Any words of wisdom on when to pick one API over the other?

Also, can we define a DataSet via the DataSet API and then later use it with the Stream API? Not clear on that...it looks from the docs like the answer is "yes", and from other forum responses that the answer is "no." The examples show a Stream including a dataSet.id property, and then a full DataSet schema definition. I'm not clear how the two sets of functionality overlap. For example, if you open a Stream to an existing DataSet, do you have to provide a schema? What if it disagrees with the existing schema? I'd like to be able to set everything up with the DataSet API and then use the Stream API for pushes, where justified. If I need to maintain schema definitions on the client side for each push, that will be helpful to know in advance of designing and writing the code. If DataSet and Stream aren't interoperable, that will be very important to know in advance.

Thank you.