IMHO, for House districts in states with metro areas that have 3-5 million+ residents, we SHOULD start by drawing multi-member districts that follow the metro area's de-facto boundaries, so an area like southeast Florida (Dade, Broward, eastern Palm Beach) would elect 3 or 4 representatives by some NON "first past the post" system, like ranked-choice multi-member districts.
Basically, a method that distributes party ratios more or less the way proportional representation would, but leaves the selection of actual candidates to voters rather than parties.
I forgot the system's name, but a few weeks ago I stumbled over a rather clever one developed by someone about 25 years ago. It's Condorcet-like, in the sense that it tends to favor non-polarizing centrist compromise candidates who are everyone's grudging second or third choice... but allows ONE polarizing-but-popular candidate from each Party to slip through IF they have rock-solid broad support among Party members... but the more polarizing they are, the more weight the system automatically throws to non-polarizing candidates for the remainder of seats.
The nice thing about a system like this is, it allows voters to gerrymand THEMSELVES in an adhoc manner each election, and decide who among the pool of winning Representatives is THEIR representative.
In the context of South Florida, such a system would almost guarantee the elections of:
* A black Democrat
* A staunch Cuban/Venezuelan Republican
* A Centrist Democrat who's pro-Israel
* A dice-roll between a very, very neutered moderate Republican, a very conservative Democrat, or a fickle, enigmatic Independent... who's nobody's first choice, and probably not many people's second choice, but almost everyone who doesn't adore one of the first 3 can view as grudgingly-tolerable. In an election where the Republican in the first 3 seats is polarizing, a Democrat or Independent would probably win. In a race where a "normal" Republican (think: Jeb) won, another normal, semi-bland Republican might win. But it would be almost impossible for polarizing Republicans despised by everyone else to win BOTH seats.
Put another way: under this system, the victory of ONE polarizing candidate automatically stacks the deck against other polarizing candidates appealing to the same base.
The catch is... only a few metro areas in the US really have the size & political diversity to pull it off successfully. Southeast Florida unquestionably does. Central Florida would be tougher, because neither "Orlando" nor "Tampa" is quite big enough on their own, and the I-4 region is just a little too big and non-cohesive to combine.
Outside of Florida, obvious other areas where it would work are Los Angeles, the SF Bay Area, Dallas, Houston, Atlanta, Chicago, Seattle, etc.
For metro areas like DC, St. Louis, Kansas City, etc., it probably wouldn't work, because their populations are split between multiple states, so you don't quite hit the 5 million+ "sweet spot" where you'd "naturally" all-but-guarantee at least one winner per major party, plus the dominant cohesive ethnic blocs.
NY could pull it off through sheer brute-force size, even if NJ & CT had to find another way. Boston is "iffy" -- if all/most of Mass. were a single multi-member district, it might work. Philly is similarly problematic. As an eastern-PA super-district, maybe... as "Philly" per se, probably not.