So you are interested in periodic events happening every 15 minute ? At such low frequencies, does it make sense to do frequency domain (1 millihertz) analysis instead of time domain analysis ?
Some GPS receivers have also higher frequency output than the 1 PPS output. Use it for the PD reference and you get much frequent OCXO output.
On some GPS modules the 1 PPS output is purely software generating having lot of jitter. Typically one cycle is little short but next cycle is longer so the average is still close to 1 s.
A good OCXO should have quite good short time stability, thus needing only disciplining the long time drift. Due to the large jitter in the
1 s signal, a large, several second (0.1 Hz or 100 mHz) integrating filtering is required between the PD and OCXO. Any external noise entering this control line will cause frequency deviations, but selecting an OCXO with low control voltage sensitivity should reduce this risk.