One other thing to keep in mind: If your "baseband" VCO (e.g., AD652) is at 10kHz, and it drifts 0.1%, it will move 10Hz. If your heterodyne output is at 10kHz, and its VCO is 10.01MHz (crystal oscillator is 10.00MHz), and the VCO drifts 0.1%, the output will move 10kHz - a 100% change! Obviously, it gets even worse at lower frequencies. Now, you could use a lower crystal frequency, but then your VCO deviation would have to be proportionally larger, which gets to be more difficult at RF frequencies. I'm not saying it's impossible (but it might be), and if you don't need to go that low in frequency then maybe it's not a problem, but it's something to think about. If the high-frequency VCO is in a feedback loop like a PLL, then these drift problems may go away, but I don't know how to do what you want with a PLL.
If you don't need instantaneous frequency agility, think about direct digital synthesis.