Linearity depends on two basic items. 1. The transfer function at the signal level you are using. 2. Any feedback in the circuit. Unfortunately there is a tradeoff between these items. JFETs are very linear but the gain is so low that you cannot use much if any feedback or the gain will go to zero dB. Bipolars have high gain and high distortion but can take a lot of feedback without reducing the gain too low.
Degenerative components improve the linearity but in eseential, linearity totally depends on the tranfer function its coefficients of active device that works under certain operating area ( biasing, circuit topology, driving level etc.
So, only increasing degenerative components will improve up to some level but beyond that level, essential nonlinear components will take its role.
There is an optimum output matching impedance for linearity of PA. But at this point the gain , the output power and the PAE are not what you want, so there is a trade-off impedance between them