Beside reducing thermal errors there is a trade off between bandwidth and Q, also an approach might favor slightly more noise but less distortion.
I personally reach for the digital control and readout for convenience without any guilt of being overly analog. My expectation for the outcome is not set so high that I cannot enjoy the project for what it is.

 

You could put an analog sine wave oscillator inside a phase locked loop. That would give you stability as good as your timing reference. If you want ±30 PPM over the commercial temperature range, you should count on keeping your reference oscillator in an oven.

It can be done, but you would have to go to some trouble and expense to do it.

My friends into VHF DX'ing keep theirs in a styrofoam cooler. It is also handy for storing the rifle scope they use for aiming their antennas.

 

If I understand correctly you want an adjustable sine frequency generator for 9.1-10.1 Mhz
you would use the LM4132 voltage reference having 20mA output and with you wanted to avoid adjusting back and forth.
Possibly the LTC1799 with adjustable frequency and supply of 3V. looking at the data sheet p.7 the Rset
value can be calculated. datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/LTC1799.pdf
Video showing the 1799 running, A 25 turn 1 M pot too sensitive. 10MHz ringing. video forwarded t=4:52
data sheet to on page 5 makes me think there is potential on a better setup than video shows should give a descent 9-10 MHz.
using correct variable Rset test it see what it needs.