Dave's redistricting only goes up to 499 districts per map...so yeah that kills the idea SMDs. If we do MMDs, the I suggest not having a set number of seats per MMD. Instead, we should seek to nest seats within cities and counties with a maximum/minimum seat count for the MMDs. Something like 5 seats minimum and 10 seats maximum per MMD with a preference to geographic barriers rather then a equity of seats.
When I was doing my cube-root legislatures, I was using a range of 3 to 5, with an exception to go up to 6 if needed to avoid going outside a county or town. A county entitled to 5.5 members could be kept in the county as a single MMD, rather than one MMD in the county, and another that included part of the county and other counties.
I was apportioning fractional representative (rounded to the nearest 1/5). This would be used with temporal weighting, where the number of representatives would vary throughout the 5 terms in a decade. A district entitled to 3.6 representatives would have 3 representatives for 3 terms, and 2 representatives the other two terms; rather than 3 representatives, an arm, a leg, and a nose of another.
You can apportion in total 5 X the nominal number of members. This must be a multiple of five for the whole country. So floor(US_Population/30000) * 5 = total 1/5ths of representatives. Apportion that many among the states based on St. Lague.
Within a state, counties with more than 15/5ths of a representative should form one or more districts, with a maximum of up to 30/5ths for any district. Smaller counties should be grouped together, but it may be necessary in a few instances to place a smaller county with a larger county, and treat the two as a whole - keeping the smaller county in one district.
After you have generated the districts, reapportion using St.Lague so that the sum of the districts equals the statewide total. This may result in a district entitled to 3.52 representatives receiving 3.4 rather than 3.6 representatives, but this is a small error.
