I think this may be going for the perfect by sacrificing the good. There is nothing inherently wrong about relying on voter patterns. It's only been wrong to us

I think this may be going for the perfect by sacrificing the good. There is nothing inherently wrong about relying on voter patterns. It's only been wrong to use voter patterns to gerrymander. One could require that a redistricting lead to a layout that averaged 0% efficiency gap for all elections since the last redistricting effort. From that point forward, voter patterns and distribution would change, but that is exactly the reason we periodically redistrict anyway. There would still be some distortions but it would probably be a marked improvement from the status quo. Redistricting without looking at voter demographics would lead to greater efficiency gaps. It's true that it would favor one party or another from state to state, but 50 states (or ~400 districts) is not enough to expect it would cancel out. And there are plenty of other good reasons to look at voter patterns - cohesive neighborhoods, historic geographic personality of districts, etc. It's more complicated than just aiming at contiguousness itself.