This is a comparison, on the same map with the same cities, of two road configurations at Renaissance. All road tiles also had an Improvement, as that is needed to maximize UA. Tile counts, minimum worker turns on Epic for all roads and improvements, culture from terrain, and population set to work those tiles tallied. This is on a Great Plains map, to easily maximize tiles the UA applies to.
Direct Routes:
I swerved mildly to run over 'high yield' hexes, but generally built straight from capital outward. Total of 40 tiles (including city tiles) benefiting from UA. Culture from Terrain - 110 in Renaissance. Roads: 32 448 worker turns minimum on Epic (160 for Roads, plus minimum 288 more to put an Improvement on each). 9 of the workable tiles are in the Capital, which is a big chunk of population that isn't being a Specialist or working a high-production tile.
Crazy roller-coaster madness:
Did a spiral through the 3rd ring around each city, linking in without crossing, ignoring what hexes I ran over to maximize sheer number. Made some very strange choices in the more mountainous terrain. Total of 104 tiles benefiting from UA. Culture from Terrain - 351 in Renaissance. 100 Roads for minimum 1400 worker turns (500 worker turns for road, 900 for other improvements). 98 population between 6 cities dedicated to maximizing the UA, or about 16 per city.
Things I've learned: Tiles with resources and Flood Plains seem to get less benefit from the UA than tiles without. This was not accounted for in route planning. Changing the direct routes to avoid resource tiles (counter-intuitive) boost Culture from Terrain to 123 and total tiles to 44. These tiles would need 532 worker turns minimum (190 roads, 342 improvemens) and use 38 population across 6 cities (6 in each, 2 extra in Capital).
I imagine the third example is the intent. By maximizing your output, you can nearly triple your rewards at nearly triple the investment in worker turns. That does seem a bit off to me. I would need to play a few full games of each approach to get a feel for how much easier this makes a Culture win. I'm also thinking that, on a more 'normal' map such as Continents or Pangaea, this sort of scenario would be somewhat rare.
As for changes, I really dislike the idea of just cutting city connections. Then the best play is to manipulate where the trade routes travel, which is even less intuitive. It also makes Arabia much more dependent than they already are on getting the right terrain and neighbor. I would much rather see something like one of the following if a change is needed:
1) Plains and Dessert tiles gain extra Food, Gold, and Culture in cities with at least 1 international land trade route (scales with Era) - You can balance on the idea that every proper tile is going to get the boost, at the cost of Arabia having completely decentralized trade routes and not being able to reliably boost more than 1 + 1 per Era cities. Trade Route granting wonders, social policies, and technology become much higher priority, as more trade routes means more cities getting the boost and maybe more routes from your primary trade hub if you've saturated your cities.
2) Same as 1, but in cities with a City Connection and not increasing by Era. More boring IMO, but easier to balance.
3) Cities with City Connections or International Land Trade Routes generate extra Food, Gold, and Culture each turn (scales with Era) - Easy to balance, removes the extra Plains and Desert flavor, potentially leaves Arabia with a ton of useless tiles if they start in a heavy Desert area. They still have Arabian Terrain Bonus for Plain and Desert flavor, and their UB offers extra land trade incentives.