I'd frankly not replace the crystal with an oscillating circuit of adjustable resonance frequency, like you're recommending, but the whole oscillator: remove (the unlabeled) C3, and just insert the output of your complete VCO circuit there, AC coupled on its own.
Just going from "the crystal has a component-wise identical equivalent circuit" to "I can replace it with that circuit" is a bit questionable: You're relying on a simplified model of a crystal be functionally identical to the simplified model of a VCO – that might in many cases work, but it might require a lot of hand-tuning, or simply might not work. The phase response of a crystal is pretty sharp, and the lag you'll see in a RLC might simply not do it for the transistor together with C1,C2.
However, VCO circuits can be built that include an active part (e.g. at least one amplifying transistor) at little added complexity, but without the hoping and tuning. So, I'd do that, verify its operation in isolation, and then have a switch that switches between the original colpitts and the VCO.
I know this is not the "market" you're in, but I'd like to also point out how frequency-agile radios do it these days to have both crystal-grade frequency accuracy and VCO-grade adjustability whilst maintaining low harmonics: They use a fixed crystal as the primary source of frequency, then use adjustable frequency dividers (i.e. fast digital counters), fixed frequency multipliers (e.g. square the sine signal, e.g. with a diode, filter out harmonics) and adjustable frequency multipliers in control loops.
The adjustable part is typically realized through having a VCO with a counter that counts the VCO cycles during reference cycles, and adjusts the VCO control voltage accordingly. For example: You have a 10 MHz crystal, you want to get a stable 88.88… MHz; you program your counter to count 9 crystal cycles until it emits one pulse at 1.11… MHz (that's 10 MHz / 9). Then your PLL is programmed to count the reference cycles in e.g. 80 VCO cycles; if that's less than 10, increase VCO voltage, if it's more, decrease.
In reality, these devices can be even more complicated, and allow for non-rational ratios of in- to output frequency through the elegance of a bit of algebra.