Splits may also create unintended resonances at some random frequencies and the result might be actually worse than without a split. So if there happens to be an aggressor signal at that frequency, it might make things worse. Having a single solid plane with proper placement of functional blocks is usually the solution with least amount of surpriseness.
Since this kind of discussion is usually just a list of everybody's opinion, I'll post measurement which I made some time ago.
Test board had an oscillator which drives a CDCV304 clock driver, which drives some 50 ohm transmission lines terminated at the ends. I made two exactly similar boards, but other one had all copper removed between via fencing:
These measurements are measured from that SMA connector connected to a Rohde & Schwarz FSV 7 spectrum analyzer. Here is a measurement from a board using continuous ground:
And here is a measurement from a board using split ground:
As you can see, there is some gain for creating the split but then, at ~580 MHz the situation becomes significantly worse due to a resonance. High power stuff might be a valid reason to do split grounds but for signal stuff, I think it is definitely not a magic bullet which solves all the grounding problems.
Regards,
Janne