Is it possible to use 2 base units to increase the accuracy of the rover?
Either feeding 2 correction streams to the rover, or feeding corrections from 1 base unit to the other base unit?
It seems that the units share data on the satellite feeds, so it seems that having more feeds for the rover would increase accuracy or reduce time to get a fixed position.
While theoretically possible, it is not something that is coded into RTKLIB right now.
Also, you must think about where you get your known locations from for both bases, because if the two base coordinates are not perfectly aligned with respect to each other, then you might be introducing more ambiguity or some kind of bias in the combined solution.
Nothing stops you from post-processing separate results from two bases and comparing the results though. This way you have your eyes on both solutions and can use this as a verification step in your process. I have done this many times.
Another point is that the base is usually in a place with good sky view, and it is the rover which must go to places with inferior sky view. Having a second base may not be that helpful to the goal of increasing accuracy, but it will increase confidence!
In the survey industry, it is often preferable to have 2 base stations and a rover. The second base forms a check on the primary measurement from the first base.
For legal surveys here, we are forced to provide a mean measurement with misclosures. For the most part, it’s a simple exercise in excel. But I have never hear of a rover calculating a fix from 2 bases simultaneously.