Do you know if there is a reason PV
farms avoid concentrated photovoltaics?
https://en.wikipedia.org/wiki/Concentrator_photovoltaics
This should allow swapping out a small PV module and more cheaply upgrade a farm as PV module efficiencies improve in many small steps.
One could argue that one would have to upgrade the inverters at every step, but that would be wasteful: just size the inverters to accomodate many future increments (if inverter lifetime is bottleneck one has to estimate future efficiencies of PV modules, if inverter lifetime is not the bottleneck just use the thermodynamic maximum efficiency to size the inverter once, and only upgrade the PV module over time).