Well, first a correction regarding no (solar powered) rover or instrument on the Moon ever making it through the night and nobody even trying. There has been several instruments and two rovers that made it through at least one, if not several lunar nights:
- Surveyor 1 was solar powered, landed on the Moon on June 2, 1966 and continued returning data through January 7, 1967.
- Surveyor 3 landed on April 20, 1967 and was active until May 3, 1967.
- Surveyor 5 from September 11, 1967 to December 17, 1967
- Surveyor 6 that landed on November 10, 1967 barely made it through a single lunar night and didn't return any useful data after it, but it managed
- Surveyor 7 that landed on January 09, 1968 also suffered battery damage during its first lunar night, but it managed to return some useful data and not just tease that it's still alive then die like its predecessor
- Lunokhod 1 landed on November 17, 1970 and was last contacted on September 14, 1971
- Lunokhod 2 landed on 15 January 1973, achieved record for off-Earth roving distance only beaten by Opportunity rover on Mars 41 years later, and it is estimated to have died due to overheating on June 3, 1973.
This doesn't tell us much about durability of today's solar panels on the surface of the Moon, or effects of thermal cycling though, since all of these achievements are probably older than your average readership here.
Luckily, NASA comes to the rescue in Analysis of Stationary, Photovoltaic-Based Surface Power System Designs at the Lunar South Pole, Joshua E. Freeh, Glenn Research Center, 2009 (PDF). Environment selected for for the study is arguably even harsher, though thermal cycling periods average out to about once per year instead of once per month closer to lunar equator. Still, it gives good overview of challenges and assumed lifetime of photovoltaic-based regenerative (using hydrogen fuel cells) power systems.
If I'm reading the paper correctly, total regenerative power system's durability is estimated to 10 years, with lowest hydrogen tank's state of charge at 10% for a 9.9 kW RFC (Regenerative Fuel Cell) option 0 with total power system mass of 2,860 kg.
One thing worth noting here is that fuel cells generate substantial amount of heat energy (more efficient hydrogen fuel cells operate at roughly 1,000 K), so thermal management could also be designed to be self-sufficient while fuel cells last. Additional problems affecting lifetime of solar array cells are also dust, radiation, and micrometeorites.
Dust problem is explained in more detail in "Kicking Up Some Dust": An Experimental Investigation Relating Lunar Dust Erosive Wear to Solar Power Loss, Jeremiah N. Mpagazehe et al., 2013 (PDF) but could be mitigated with, say, Optimization of Photovoltaic Performance through the Integration of Electrodynamic Dust Shield Layers, S. Nason et al., 2014 (PDF) and strategic placement of arrays where they wouldn't be a subject to bombardment of kicked up dust. Radiation environment is somewhat harsher than for solar arrays in GEO due to secondary reflected off the surface radiation, so PV durability would likely be closer to those of GPS satellites orbiting within the Van Allen radiation belts and I'd assume ~ 2% degradation per year instead of more average 1%, tho that's my own assumptions, and micrometeorites are something you'll have to design to live with. Certain PV designs (like, say, multi-junction cells that NASA loves to use for its probes and orbiters despite larger mass than, say, thin film PVs) will tolerate single cell defects better than others, but you'll want to design with some reasonable margin and redundancy in mind.
If I had to design what you're doing, I'd probably go with Stretched Lens Array Squarerigger (SLASR) and state-of-the-art multi-junction cell design. For example, one proposed Europa Orbiter design described in Solar Power for Outer Planets Study (PDF) uses Advanced SLASR Arrays SOA MJ Cell at 45 kW BOL at 1 AU, using four 7.0 m diameter Ultraflex arrays at 513 kg total mass. And that was in 2007, you can assume that state-of-the-art somewhat improved by now.
Some other documents worth considering:
And of course NASA Technical Reports Server (NTRS) is your friend whenever you're looking for information on something that might have been considered before.