[3Pi]

It's worth noting that the SpaceX satellites only have a mission length of 5-7 years, and the Spectrolab solar panels are built to last for a typical 15 length mission (from the link you provided). Perhaps they will opt for (presumably) cheaper components given the shorter lifetime.

[walrus01]

15 years is the nominal design lifetime for an "insured" geostationary telecom satellite... When SES or Intelsat launches a new 3500 to 6000 kilogram sized satellite it's insured by a third party company. There's specialists in this. I'd say it's much more likely the SpaceX program will be self-insured so they can take a risk with satellite bus technology and other tech that has never been flown before.

How this will affect their choice of PV cells I don't know.

With 4000 satellites it's likely that their approach to each satellite will be smaller and much less redundancy:

If a traditional 5000 kg geostationary telecom satellite can be compared in analogy to a big, expensive, 4U, quad socket xeon server that has multiply redundant everything.

These small satellites might be much more like a facebook open compute blade server, 1+0 and redundant nothing, but much, much cheaper to build and with a software architecture tolerant of entire nodes failing and disappearing from the network.