PD evasion by "beaming" and why it still works?

I’ve been thinking why the “beaming manoeuvre” (turning perpendicular to the scanning aircraft to deny Doppler shift with constant distance) used to evade modern fighter radars is still considered to be so effective.

I’m quite at a loss to explain, how modern high CPU capacity radars aren’t provided with theoretically pretty trivial ways to counter it. Or are they?

The principle of using Doppler filters is sound against LD targets, because without the filter you wouldn’t be able to discern a beaming target anyway (except with additional trouble, back to that later), but in look up situations against a clear sky, the Doppler filter removes information that the radar could use to display a target!

The only radar that I know of to have a Doppler filter-disabling mode available manually is the AWG-9 (called PS – pulse search mode). This brings out two questions:
a) are there any other PD radars out there that have this functionality
b) if yes, are any of those radars able to take advantage of this automatically as opposed to manually

I mean, for a digital radar, it doesn’t require much extra work to just display the target without the information from the Doppler shift, if the logic can decide based on absence of ground clutter (later about how to determine this) that the target is above horizon. Sure, you lose speed information, but if the radar has a tracking file on the target, it should still be able to both present it with guesses extrapolated from earlier better data and also be able to feed this new data (even though lesser quality) as targeting information to weapons.

Also, I humbly submit, that with sufficient computing power, available for today’s radars, they should be able to do much of the same trick in LD situations, being able to pick up targets from the ground clutter even when the target is below the DS threshold.

One way is to use range information in combination with sorting out the ground returns heuristically. The radar could use aircraft navigation (attitude+altitude, 3D coordinates) data to determine where the horizon line is in relation to the return distribution and then compare each return value in the same “earth relative horizontal line” to each other and when there is enough returns (this sets the “average altitude of ground”), judge that “this is ground” and drop it from further analysis. However, any returns that have arrived earlier from the same “bar” scanned than the ground returns, must be airborne targets and so worth being used, even if their DS is below threshold.

The above method should be able to detect any beaming LD targets above the “average altitude” (i.e. targets below highest terrain features + range ambiguity), but would still leave very low flying targets undetectable without enough DS. By investing a little more computation, it should be possible to weed out even those, if close enough.

To do this we basically need to handle the ground returns just as GMT mode, albeit with inverse speed threshold (we eliminate anything that moves slower than possible airborne targets). Sure, the signal we send is not optimized for this, so we would get much reduced range. This could be compensated my mode mixing, spending some of the time scanning low elevations in a more “GMT like” mode and some in AA modes. Probably the most economical strategy would be to employ “GMT like” only when the radar loses a LD target in conditions where it was able to track it with the Doppler filter on.

Is there any sense in what I’m proposing? Are they already doing this (let’s say with AN/APG-77), but we just don’t know?