The Oncore VP has a 9.54 MHz oscillator and thus exhibits a sawtooth of approximately ± 52 ns (1 / 9.54 MHz = 104 ns).
Below is a 10 minute trace of the 1PPS output of a VP (log30177.txt). The data was obtained with a 53132A TIC and there is a deliberate ~150 ns offset in the 1PPS reference (to avoid negative numbers).
Take a look at the plot and observe the following:
Now, here's the cool part. Near sample 480 there is a slow ramp, the kind of sawtooth trend that is not ideal for GPS time interval averaging. What happened right at sample 490? Suddenly the slow, GPSDO-unfriendly ramp turned into a fast, GPSDO-friendly sawtooth. Why?
The answer is that I was watching the sawtooth go into this slow ramp starting around second 420 and a minute later, I aimed a hair dryer at the bare Oncore VP from a meter away for a few seconds. Bingo! The slow ramp goes away and the sawtooth gets good again.
If one were to do this for real a small resistor mounted on top of the XO would work much better. The idea is to monitor the phase, or the phase differences, and when a slow ramp is detected (which can be done within, say, ten seconds), a pulse of power sent to the resistor to temporarily heat the XO, to break the close synchronization between GPS and the XO.
10 minute VP trace #2
10 minute VP trace #3, showing some nice hanging bridges: