All:
Good news, below is an infinite persistence plot of the audio out of two Voter2s in simulcast mode, each running from their own crystal, GPS receiver, and antenna. The top screen is just to show the pulse positioning, a test mode in the Voter2 generates this pulse every 80ms (4 packets). I have another mode that does a pulse every 11 packets just in case I’m being fooled with an error of exactly 80ms. The lower picture is zoomed in on that pulse. At 500us/div there are about 4 audio samples per division.
As you can see the audio from the two Voter2s are always within 1 audio sample of each other. My current code is only trying to sync up to whole audio samples, so this makes sense. I have a pretty good idea on how to enhance that accuracy to less than an audio sample. While not sure, I feel even if I can’t get better than this, that should be pretty good for simulcast.
With that I am confident I won’t need the 10MHz reference input now, so feeling the hardware design is pretty stable. Back to fixing software bugs and missing features.
will (N0XGA)
p.s. One picture limit? Just combine both pictures into one image! 
I fully support newbie limits, it protects us all.