I would imagine such an experiment having at least 4 devided into 2 groups of 2 - with each group conraining a pi experiencing normal use and another experiencing extraordinarily high use. One group could have heatsinks and the other not.
More is better but possibly impractical.
Could be crowd funded with each donator entering a sweepstake to bet on the expirey point.
Bit of a macabre fantasy really, but would be interesting to see how much this little board could take.
I guess this sort of testing should have (and possibly has) been done by the foundation prior to release any how.
Ostendo ignarus addo scientia.