@Naeven
The goal is for everyone to get a few policies in most trees, and most policies in a few trees. The maintenance cost reduction is designed to be one of the policies wide players want in the Tradition tree. If tradition was all-tall and liberty all-wide, we wouldn't really have a choice. It would be an automatic decision.
@Naeven, jacktannery
I agree that Korea and Babylon are probably equally good overall. Building farms reduces our gold/production to actually
build the things we research with all that science. It's got downsides.
@mitsho
The main thing affecting early barbarian pressure is the turn when barbarians can enter civilized territory. I'd be okay with reducing it. The current settings are:
Barbarian release turn
100 Settler
90 Chieftain
80 Warlord
70 Prince
60 King
50 Emperor
40 Immortal
30 Deity
How about this change:
100 Settler
80 Chieftain
65 Warlord
50 Prince
40 King
30 Emperor
25 Immortal
20 Deity
@Bernd-das-Brot
In v1.11 compass requires optics, currency, and philosophy. I can remove the philosophy requirement if we want. This would reduce the number of techs needed for compass by only 1, since philosophy and currency both require drama. This would not significantly affect gameplay, but might feel more realistic, and would make that part of the tech tree a little easier to read.
Land and sea units are mutually exclusive opposites. We have limited economic supply for units (gold/production/resources etc), so we must choose between land and sea for each unit. In games with lots of water, we build fewer land units, and vice versa. Everyone needs civilian techs, so civilian techs are in the middle of the tree between the opposites of land and sea. If sea units were in the middle, it would cause problems for conquerors on land-focused maps, since people would have to get sea techs we don't need to reach the civilian techs we do need.
@jacktannery
Thank you. Based on the available information, the Korean UA increases our average tech rate
15-20% 
. Here are some UA comparisons:
|
Early
|
Late
|
Rate
Korea
|20%|15%|

research
Arabia
|10%|40%|

gold
Rome
|15%|20%|

production
This is the average amount each UA increases a national rate of techs, gold, or production. The midgame data you provided helped a lot because midgame is harder to theorize than early game, since there's more variables to consider. In the midgame, farms increase jacktannery's science rate about 15% in both capital and satellite cities, which matches my own experience.
A late-ancient empire is easier to analyze. If we have a 2-city empire where each city has 6 pop, 5 farms, and a library:
12 from national bonus
12 from population
10 from farms
22 from buildings
4 from specialists
------------
60

total
16% from farms
An early-ancient game with a 2-pop capital and hall:
12 from national bonus
2 from population
4 from farms
2 from buildings
------------
20

total
20% from farms
Notice how the national bonus balances the value of farms in the super-early game. The farms would normally double our science rate, but instead contribute only 20%. The contribution from farms remains in the 10-20% range throughout the game.
I will reduce the strength of hwach'a, since it seems everyone agrees more about that unit. This brings our G&K-GEM comparison to:
G&K: 150%

unique caravel restricted to coast
GEM: removed
G&K: 180%

unique trebuchet with -25% vs cities
GEM:
140%

unique trebuchet with -50% vs cities (was 180%, -50%)
G&K: 4

specialists from the start of the game
GEM: 2

specialists with a national wonder
G&K: 200

to 6000

instantly from science buildings and wonders in the capital.
EM: 2

on farms.