The problem is really that the FFT function is insisting on filling its 128k FFT points, even if that takes a very long time and the actual FFT window size is much much smaller than that.
The benefit is that you can change window side and position after capture, which is normally nice, but not worth the downside in low-freq applications.
It would be very nice if instead you could simply force the capture size to be the much smaller window size. But I have a suspicion that the reason this is not changed, is that that could require a larger-than-tiny change of the FPGA logic, rather than just a software change.