Dave:
The classic analog spectrum analyzer design is multiple conversion - with an up-converting input stage.
The input runs through a low pass filter, attenuator, etc. and then is up-converted to a frequency above the highest input frequency. For a 1.5 GHz high limit, the 1st IF might be in the 1.8 to 2 GHz range. A rather wide fixed bandwidth bandpass filter is used after the 1st mixer. Up-conversion is used for several reasons, one of which being that if the 1st IF frequency is someplace inside the operating frequency range you have a major signal leakage problem. Another one being that it's much easier to build a LO that spans 1 octave, say 2-3.5 GHz where the 1st IF is 2 GHz, than to build a LO that covers 9 KHz to 1.5 GHz. And, of course, image rejection becomes difficult with a lower frequency 1st IF. (Some of these problems can be solved with digital processing and I-Q samples, but there are still excellent reasons to go with an up-converting input design.)
In your tear down, the 1st IF bandpass filter is likely the one formed by the |unun| shaped traces. Probably 200 MHz or so bandwidth, centered at the 1st IF.
In a classic analog spectrum analyzer, this high IF is then amplified and run to a second mixer where it is down converted - perhaps to an IF around 300 MHz where some very wide resolution bandwidth filters can be used, and then a third conversion to a much lower frequency such as 21.4 MHz where crystal filters with resolution in the tens of Hz or less are feasible.
Looking at your video, the design seems to follow the classic analog approach with up-conversion to the 2 GHz range or so, followed by a second mixer / LO to a lower frequency. Then into the ADC for the rest of the signal processing.
Jack