Several ideas about supplying the palaces with foods and goods were bouncing around my head. I don't remember the order I put them together, but the following might be close.
The requirements for supplying foods and for supplying non-food goods to palaces are very different: a palace will start to devolve 1.5 to 2 months after last being topped off in furniture or oil or wine (or 3.5 to 4 months after last being topped off in pottery), but a full palace will start to devolve 23 or 24 months after starting to want more food (and not getting it). Also, the demands for foods and non-food goods by markets are very different: the market(s) serving 19 palaces need to resupply with wine every 2.63 months on average and with pottery or furniture or oil every 5.26 months on average, but a market needs to be resupplied with food much more often since topping off a food in a full luxury palace empties a market of that food. So, it's important for supplying non-food goods that a market's traders pass by palaces frequently (which is why I used 16 gatehouses in LuxPal256), but it's more important for supplying food that market buyers get food from granaries quickly.
I've liked the idea of separately supplying food and goods for many years. (My Lugdunum Cliff Dwellers did that for its grand insulae, for a different reason: food was more easily available on the high side of the cliffs but oil was more easily available on the low side of the cliffs.) In addition to the reason give in the previous paragraph, the separation prevents the problems with non-food goods that a granary running low on a food can cause (shown in reply #174). If I could use only 1 or 2 markets to supply non-food goods to a palace block, then I could cut the number of gatehouses in half, using 1 gatehouse for the non-food goods markets for 2 blocks and not using gatehouses for the food markets.
But there's a problem with having 1 or 2 non-food goods markets per palace block. Well over a month will elapse between walks by market traders, so a market may have to distribute 57 pottery and 57 furniture and 57 oil and 114 wine on 1 pass through a 19-palace block. (The amounts would be greater if over 1.5 months elapses between walks.) Unless they can resupply very quickly, even 2 markets may run out before supplying wine to 19 palaces. So warehouses would need to be very close to non-food goods markets, and I obviously didn't want to increase the number of warehouses (at least, not much). Using 2 getting warehouses (each getting 2 non-food goods) in 6 places and using 1 such getting warehouse plus 2 accepting warehouses in 2 places might work, with only 2 extra warehouses (which might be acceptable with sufficient building savings elsewhere). I started to check whether a getting warehouse could get wine and another good fast enough from the most distant location, but the ideas kept coming.
If the number of palaces per block was increased from 19 to 20, consumption of non-food goods would exactly equal production on average. (16 more palaces would mean higher food requirements and more fountains (the 64 palace fountains of LuxPal256) and problems with school coverage in some orientations, but I had managed to solve a similar (perhaps half as bad) school problem when increasing the number of palaces per block from 16 to 19, so I put those problems aside.) 2 warehouses with 2 non-food goods each ought to work fine with local production if consumption equals production on average, so there would be no increase in warehouses. There would be 8 non-food goods supply areas, each supplying 2 blocks of 20 palaces, each with 2 pottery workshops, 2 furniture workshops, 2 oil workshops, 4 wine workshops, and appropriate raw material producers.
(Around this point I realized that even if consumption of non-food goods did not equal production on average, the warehouses could be adjusted during development and should work for several years without intervention. I could have cut the number of warehouses in LuxPal256 in half, as suggested by Trium in reply #166.)
Assuming that a non-food goods market never runs out of a good, when a palace is topped off the amounts of pottery and furniture and oil supplied are equal and the amount of wine supplied is twice as much. It should be possible (taking care during development) to have each non-food goods market run low on pottery and furniture and oil at quite different times. It should also be possible to have wine run low first if both wine and another non-food good run low at similar times. With warehouses quite close, 1 market should be able to resupply with wine and another non-food good before it runs out of either, if wine runs low first. 1 non-food goods market per palace block! I'm assuming that a market that doesn't get food and is low on exactly 1 non-food good will send its buyer to get that good within 1/16 month--if it sometimes doesn't do that (I ran into a similar but worse problem in Pharaoh) this may not work, but I haven't seen any evidence of such a problem.
With only 2 markets taking goods from a set of warehouses, and with consumption equaling production on average, only 1 warehouse should be enough. The warehouse could still get somewhat fragmented, but if pottery and furniture and oil each sometimes gets down to 0 or 1 in the warehouse then it might never become a problem. Only 8 warehouses!
But wait, there's more! With 2 markets and 1 warehouse, the warehouse pulley could be used instead of a gatehouse. (This would be troublesome in some orientations, but I put that aside too.) The walk targets for the markets would have to be planned to prevent the market trader from going through the pulley and wandering around industrial roads, like the fixes I made in LuxPal256 for a design flaw described in reply #154. No gatehouses! (Also, the market buyer with basket boys would appear for only a moment.)
Long enough post. Food later.
[This message has been edited by Brugle (edited 12-20-2010 @ 02:20 PM).]