The question isn't really whether the radalt fails in any particular test (spoiler: it probably won't). It's whether there's an edge case where the safety case

The question isn't really whether the radalt fails in any particular test (spoiler: it probably won't). It's whether there's an edge case where the safety case for the aircraft depends on correct operation of the radalt AND there's a failure. I haven't seen the fault tree for the 737, of course, so I don't know what reliability is required from the radalt to be safe; but with the radalts I've worked with it's not unreasonable to have an undetected failure rate requirement of 10^-8/hour or better. (Detected failure rates can be much higher.) Testing enough to show that you're not violating that is /hard/. Speaking a bit from my own experience, I suspect the guard bands on most radalts out there are wide enough that 5G interference isn't a practical issue. I've seen radalts with 30 MHz guard bands around the 4.1 GHz / 4.3 GHz limit frequencies of interest... and that's friggin' huge. (I'm also sure there are radalts with much smaller guard bands that use more of the available bandwidth for better performance.) But this is hard to analyze fully. Taking the example of the unit I'm thinking of, the center frequency is from a non-temperature-compensated oscillator -- so your worst case is an extreme temperature going to assymetric guard bands, a unit with a weak bandpass to begin with, and a 5G system pouring a lot of energy at (or even, in violation of spec, a bit beyond) the edge frequency. Probably fine. 10^8 fine?