What I don't like about this analysis is that it enshrines the notion of political parties. It makes individuals into automata that simply vote for whatever candidate is thrown at them.
The ideal is that candidates can rest on their own merits, with the party mechanic existing to provide support to candidates who nominally align with the party's interests. People vote for candidates, not parties.
Even the reality is that people are not determined by their party affiliation; New York City, for example, has repeatedly elected Republican mayors despite being overwhelmingly Democratic in state and national elections. This has even been used in a derogatory sense, as "Rockefeller Republicans" or "New York Republicans"; people affiliated with the national party but who skew to the preferences of voters in their region.
In your hypothetical state, once you created the districts, then the competition inside that district is between candidates who try to align themselves with that region; people who, for example, can be a "green-republican" to draw voters from both blocks, or just a person who speaks to the needs of voters in that district, regardless of party affiliation.
Using some sort of state-wide rank-preferential or approval system means that we remove a lot of the strategizing on the part of voters, which is a positive, but the strategizing on the part of political parties and individuals seeking office become much more complex, and not in a way that serves to present candidates that appeal to issues confronted by groups of people regardless of their party affiliation. Geography is just a heuristic for this, but not the worst one you could think of.