The data lover in me sees an opportunity (admittedly, highly unfeasible as yet another ask on their time) to ask tournament organizers if they would track scenario chosen, hours played, and if it ended on VC (which turn), concession (which turn), or on scenario end. This could be provided by the players when reporting results.
Over time, tournament scenarios would provide enough data that could be extrapolated based on S/G/A, map size, VC etc. to roughly correlate a playing time. As others have mentioned, this is still highly dependent on player speed and one could make arguments for/against tournament players as a subset of the larger player base's playing speed. Still, that dataset could form a basis for refining a rough/approximate/hopeful formula for playing time.
It could also be an ask of players when recording a play on the Archive. If known, provide playing time (H:M), which would build up a second source of data.
I liked Jim's threading the needle comparison, it's very challenging given the mix of objective and subjective data. Still, all the perspectives have been enlightening - improvement only comes from dialogue and constructive feedback.