Feedback following RAGL S11
I always find it a fascinating task to apply tournament systems from one game/sports to another and see how it pans out. It's easy to consider everything was already invented before you, you just need to re-shuffle correct things in correct order to get what you need.
In this regard, I've thought to somehow introduce BroodWar's ASL double tournament brackets into OpenRA, but it only actually works for proper multi-day TV tournaments (and it's quite ideal for that format).
What you were talking about reminded me of several systems that might be mashed up together to get the useful scenario for us. (while describing them I've realized first two are of no real use here, so I've removed the text).
McMahon system: this one I tried to grasp while playing Go online, it's used everywhere in place of Swiss, and it's the best system when there is a big number of entrees for which there is already a general understanding of skill ranking.
Basically, all entries are sorted by their general rating (in whatever system it is measured) and then split into groups, with Group 1 consisting of top players and Group N having lowest ranked. Players are then assigned starting points, players in Group 1 having the most and Group N - the least. The actual sizing of groups and point values both depend on number of players, proposed tournament length and rating gap between players.
Players compete inside groups in normal RR fashion, adding points to their tally. After all games are played, the new McMahon round is declared, and groups are reshuffled, only this time players' current point tallies are used as rank. So players who played well in their group are moved up, while players who didn't perform - are demoted into lower groups.
Number of rounds is also a subject to determine, but I assume it should at least be mathematically possible for a player in the lowest group to get into the top group by the end of the season.
The benefit for OpenRA league here is that it's quite flexible in terms of in-season forfeits (such players will just not be drafted in subsequent rounds) and rankings mismatch (players with a settled rating will be matched against similarly skilled opponents from the start, while for players with uncertain rating or no rating whatsoever there will be a round or two in lower groups, but if they severely outrank everyone there they will aggregate a lot of points and get into top group eventually). It even may reduce the forfeit count that depends on skill mismatch, because overachievers will quickly rise through groups, and lower groups will retain players of similar skill in the end.
This all comes with drawbacks of course. First, there are at least 3 variables than are not easy to decide on (number of groups, starting points per group, number of rounds). Also, don't really know whether it's a problem, but in general players of similar skill and performance may stay in the same group through all rounds, so will end up playing each other several times through the season. And it completely deconstructs the whole masters/minions/recruits system, making it a single division and allowing everyone to reach the top regardless of starting point (which is unlikely to happen, but not impossible).
Upd: More info on McMahon pairing, also includes links to some automation tools https://senseis.xmp.net/?McMahonPairing