We had a successful 4p playtest of v17 on Sunday that was designed to assess two main changes: stretching out the Unrest track, and adding "trade route" pieces (Settlers roads placed on the border between two territories). Both were successful and well-received.
The trade routes created some terminology ambiguity, but functionally they had at least two nice effects. (a) Since you place a trade route when you pass one of the two actions you're usually authorized to make, it adds a nice decision point as well as gives you something useful to do even when you don't want to take both actions. (b) It separates the scoring from "trade routes" (which now pertains to how many "star spaces" your trade routes pass through) from receiving achievement tokens from "trade routes" (which pertain to how many foreign cities your trade routes connect to). Previously, the scoring and tokens were the same and pertained to how many foreign cities your own cities were adjacent to. In practice, this resulted in big, explosive trade route networks that were pretty homogeneous across all players. This new scheme will create more differentiation from those who want to score in this category and those who want to set up to get tokens.
Several helpful changes were suggested, including getting rid of the production tracks and just calculating your production at the start of each turn, and removing the +2 Unrest penalty for failing to record a Chronicle.
The "problem" with the game is that it's still long. It took us about 4 hours for a 3 epoch game, and I suspect the last Epoch would have added another 1.5 hours or so. That's still just a bit too long, and nearly all of the bureaucracy-induced length has been cut out of the game by now. There aren't a ton of decision points, but there are many options at each decision point, so it just takes players time to evaluate their course of action. Players can plan their actions on other players' turns, but this doesn't always happen.
So, I have two possible ideas, both revolving around players' selecting their actions simultaneously. The motivation is simple: if each player takes about 3 minutes to take his turn, then if you could have all of that decision time happening simultaneously, you could probably cut the game length by a factor of 2 or 3. There's some risk that one player's actions would disrupt another player's planned actions, but a rule to let you change your planned action (maybe at the cost of +1 Unrest?) should be able to handle this; most actions don't directly interfere with other players, so this should be a somewhat infrequent occurence.
The first is probably more of a suggestion than a hard rules change. Players would write on a piece of paper which two abilities they plan to use during their turn, and what specifically they plan to do with each -- eg where they plan to build, or where they plan to attack, or whatever. Then, when their turn comes up, they simply do what the paper says. The down side of this approach is that it interfaces somewhat clunkily with the existing action selection board, with its 3x3 grid of actions.
The second approach is to (once again) rebuild the action selection mechanic from scratch, but retain the core principle of the current scheme, which is that you get to take two actions each turn. Each player would get 12 cards representing his available actions, and each turn, players would place 2 abilities on the table in front of them simultaneously. If, on a subsequent turn in the same generation, you want to reuse a card you've already used earlier in the generation, you have to gain 1 Unrest. And, the "Produce" action has a built in Unrest, while others have a built-in achievement token. (Maybe you also take an Unrest if you use two actions that each pay out a token).
Probably, some of the actions would be simplified -- eg Advance and Build would just be "play a single card/structure" instead of "as many as you want and can afford", so that there's less of a problem remembering what you planned to do once your turn comes up. Maybe you even write out your intended action, as suggested in the above scheme. (Maybe you're provided a white board to write this on).
The upside of this is that it gives you even more flexibility than the current scheme.
The down side is that it slightly disincentivizes specialization in a way the current version doesn't. But that may not be so bad. A bigger down side is that it gives players explosively more options, since 12 x 11 = 132 combos (!) are available. Maybe to reduce this somewhat there could be 6 double-sided cards, with one action on each side; but this reduces the number of available choices, but not necessarily the complexity, since you still need to remember which cards can potentially be played with other cards.
This explosion in options is a big concern: the current board has many options, but few enough that a new player can simply pick one each turn and go with it. This explosion in options could lead to paralysis, since new players may not immediately realize which actions naturally go together, something the board originally scripted for them.
I like this second idea enough to solo test it (I have no option to the first option but it's not any different in a solo test than just playing out the players' actions), but its true test will be how it holds up with newer players.
(*) One for each of the 9 current actions, 2 for "add a trade route", and 1 "x2", letting you take an action twice.