about the frame rate, i am sure we can get beyond the specified limits.
but the frame rate does definitely not scale (inverse) linear with the number of pixel lines.
exposing and reading out the sensor is split in a few separate tasks, of which some are not related to the pixel count.
first you have to discharge the sensels by asserting some circuitry, then you have to start exposure by releasing the discharge line.
discharging is important, else you will get a lot more noise, or even integrate over time.
exposing the sensels can be reduced by increasing ISO (amplification) or providing more light.
reading out is the bottleneck for huge sensor areas, usually the width doesnt matter that much, the window height is the most expensive.
i am very interested in getting something like MLV ported to your Axiom platform, making it a "common" open source standard for raw video.
especially high fps modes raised my attention