I do not think that bonding to the roof is an assured heat-sink unless what you've used to do the bonding with is thermally conductive. It would shock me immensely to learn that all VHB tapes are good thermal conductors. I'd had a couple "object lessons" in lack of thermal conduction. Substances that were assumed to be good conductors turned out to not be all that good, or even good enough. In one case that was an AAA moment as there was no self recovery possible (it let the smoke out of an ignition module with no spare on board). Made me pretty late to work that day.
Say that you've used a version of VHB that has some thickness to it. Perfectly valid choice to make as it will tolerate the two surfaces not being exactly flat better than any of the extremely thin versions. Except that this thickness comes in the form of a foam. Not exactly a good conductor when foam is more generally employed as an insulator! No doubt there are thermally conductive foams, but you can't assume that the foam used in VHB tape is that type. Then consider that this thicker VHB tape has created a small air gap between the backside of the panel and the roof skin. This gap will likely vary randomly from the tape thickness down to zero. A big part of why a large air gap (as compared to the tiny air gap created by the tape) works is because of an air exchange. The air heated by the backside of the PV panel expands and flows out from under the PV panel and is replaced by cooler ambient air. This is exactly the "cooling" process employed by Land Rover "Safari Top"'s. The air gap acts to exchange heated air with ambient creating a self-renewing form of insulation to keep the cabin cooler. The small air gap created by a thick VHB tape will act more as a thermal insulator than as a conductor of heat.
Even a coating of paint impedes thermal conduction, and the more layers of different materials there are in the conduction path the slower the thermal conduction rate. Can think of it like a series of energy conversions where you loose efficiency with each conversion. Components that are designed to be heat-sinks are deliberately bare metal, and the components attached to them are usually bare metal as well. There usually is a thermally conductive grease or adhesive smeared between them. It's purpose it to eliminate any tiny air gap(s) that might exist so as to maximize the contact surface and make the conduction as efficient as possible.
So bonding PV's to the roof skin might serve as a bit of a heat sink, but the common way that it has been described as being done on this forum is a vastly inefficient heat-sink. There is a lot of performance being left on the table. Would it be worth the trouble to address this inefficiency and increase the thermal dissipation of the PV panels? I have no idea. I've never done such an experiment. I simply go as far down that path as I'm willing to go and accept the results. I expect that others will have different points where they stop chasing the performance details (ROTI) and move on.
