Hi,
As a belt-and-braces kind of guy, I wonder whether all 3 together might work well.
I'm also wondering whether there might be a much more sophisticated approach to maintaining a constant thermal environment. I don't have all the pieces figured out in my head yet, but I'm thinking the level of sophistication of a good filter design (but in the thermal domain) where the 'input signal' is the lab temperature environment and the objective is v low pass filter giving near 'thermal DC'. As EE-types, we have a tendency to bodge a bit of arbitrary plastic / foam / copper foil around a voltage ref and call it a day. But I'm imagining a modelled thermal environment with 3D printed / CNCed / lasercut materials chosen with the right characteristics to maintain optimum minimal temperature variation with fairly low power. A basic start would be say picking a temperature for the outside of the ref (maybe halfway from the die temp to max ambient) and designing the right thermal environment with some sort of iso-thermal case surrounding the key electronics - and then that surrounded by the right insulation to get the appropriate thermal heat flow out together with a means of targeting the temperature of the iso-thermal case. Am I making any sense?
Alan