@User Q of inductors certainly is quite variable even with temperature let alone process, and the most difficult thing about them to model. In the past I have put different metal thicknesses into FastHenry corresponding to the process corners for the metal sheet resistivity and also different volume resistivity corresponding to temperature. "Skin effect" has the fortunate property that the RF resistance of the trace varies less than the DC resistance, because the middle part of the conductor is not so effective anyway. Also the skin depth increases when the resistivity increases, partially offsetting it. Circuits such as VCOs usually will need to monitor oscillation amplitude and correct the bias to account for Q variation. Anyway, until a design gets into serious mass production, the process variation is unlikely to spoil our fun, and I think the bigger problem is making a good initial prediction of Q rather than lot-to-lot variation, so for a given inductor style we can take out a lot of the uncertainty once somebody gets some chips back and characterises them. As I mentioned before, to do that, the easiest way I know of is measuring the startup current of a VCO, and at that same current measuring the gm of some copies of the same type of transistors that are used as the core transistors of the VCO, then tweaking the inductor model's Q until the simulated VCO starts at the same current (and transistor gm) as the real one. Trying to probe inductors with a VNA seems very error-prone by comparison. For such a poorly modelled device I find it interesing that the inductance value itself is probably the most tightly repeatable electrical parameter on any IC process, because a >200um diameter piece of metal will be accurate in dimensions within a fraction of 1%, whereas we would be lucky to get +/-20% tolerance guaranteed for resistors and capacitors etc. I really hope I get time to design something, preferably with a VCO in it, in which case I will use FastCap/FastHenry.