We'd need long term statistics to fully develop a model of risk. I.e., (admittedly, anecdotally) within the first year or two of having his monocrystalline setup, Dave[1] suffered damage to one panel (aerial debris) requiring a service call to replace the unit. If this happens on a semi-frequent basis (i.e. one panel per ~3 years per install) the risk profile changes not insignificantly.
I've had workmen on my roof for damaged slates due to high wind, so again, with a Tesla roof this could conceivably be a negative death rate of the solar tiles are stronger than standard tiles, which they claim they are.
