Interesting take, but it assumes a high margin of error in the voting preferences of certain constituents, which may not be true, which likely makes it not self-limiting enough.
It assumes the opposition can sway some percentage of the population in a wave in one case (maximizing seats) or that they can put up a few really good candidates that can win on personal attributes and policy, not policy alone, in the other (maximizing seat safety).
We've certainly had wave elections, have we not? 1920, 1932, 1994, 2006, 2010... They're getting more frequent, perhaps because redistricting may not be maximizing seat safety as much? (I'm not sure, I've not and will not spend my cycles on analyzing redistricting history.)
