The questions facing us for the DOT050V variant centered solely around whether the high precision regulator would make any difference compared to driving the DAC reference from the oscillator supply rail.

Red is with the regulator, blue is without.

Now, the blue has a big hump in it caused by the slow time constant of the PLL being too slow. The DOT050V isn't stable enough for a 200 second time constant, and when it switches from slow to fast to recover, there's a ~1E-9 spike and correction that has to happen. I'm not sure what's up with that. The math looks correct to me. I can only guess that the P term gets accentuated influence due to the reduced time constant. If you remove those spikes from the dataset the ADEV above 50 seconds comes down, but below that it's the same.

I think the extra couple points worth of short term ADEV are worth it. It flies in the face of Connor Winfield's advice, but I think at this point I'm willing to assert that if you read their application note, their assumption is that you trim the oscillator at one temperature and that the variable temperature compensation for the two temperature domains causes more drift than the smaller tempo could explain. But for a GPSDO, the GPS discipline represents a continuous correction rather than their example of fixed one-time tuning. That means that we can count on temperature variations being elided out over time.

And if the difference is that measurable for the DOT variant, it for sure will have an impact on the OH300. The only question left is whether or not the added LC filter on the oscillator supply does any good or not. The answer to that will have to wait until the prototype wears in a little more.