I would add these two extra joints to optimise the route some way further, to allow skips for trains that need them. Also there is a possibility to add a big way around to the left, but that seems like a big amount of rail for just a small decrease in distance for some trains
SirLightXXVII
Recent community posts
Attempting to calculate the theoretical maximum score. I think the world is a cylinder, so I will use the formula circumference times height to calculate the surface area.
Circumference we can measure by following a circle and find out when we touch our tail. To not rely on eyeballing a circle we can use the land barriers that are pretty circular. It happens soon to the "end of the game" at ~35 apples. Height we can find by measuring when we touch our own tail by going perfectly centripetal. It happens at ~138 apples. The snake is built from circles, let's assume of radius 0.5/pi, the snake starts with 7 circles and gain a single 1 each time we eat an apple.
Thus, circumference is equal to 42, height to 145, area to 6090. Area of a single circle is 0.25/pi, thus the a maximum score is 2430pi or ~76529. As this measurement cannot measure more precise than 1 circle, the error bars are 42±1 and 145±1 for dimensions, therefore the area is ±187 and the score should be ±748pi which is ±2350.

Not that close I'd say. I've managed to get to here (also grey rock is more appetite efficient as it takes only 640 appetite per white gem as compared to 900 from a metal)
There was 1 gem from eating all bring and grass, and two more from spending the appetite. To merge two greens we therefore need one more blue gem, and that is a lot; it just appears to look much less due to all of these compressions

I've managed this score, lower I think due to inefficent gem placement next to this east wall. So I'd think the 5088 is indeed the maximum, maybe can be pushed up to 5090 but not much higher. The main source that can be optimized is this upper-left puzzlebox

I have managed to extract everything from it save for these three grass tiles. Nothing is theoretically stopping it from having a better solution. I have then merged the 1-gem in the corner with a leftover from somewhere else
https://imgur.com/a/PDjOJ4m (There is one unmarked grass tile in the bottom right)
It appears that there are in total
27+21+22+
30+68+82+
185+30+96 = 561
brick wall tiles and
39+56+101+
28+62+45+
0+58+34 = 423
grass tiles. These total to 5472 hunger.
In addition to these, there are
33 tier 1 gems, 4 tier 2 gems, 1 tier 3 gems, 3 tier 4 gems, 6 tier 5 gems, 1 tier 6 gem, 2 gems of tier 7 and 8 and 1 more of tiers 9 through 12.
When we merge gems we get the following payouts
1,2,4,8,14,23,32,41,50,... and so on, increasing by 9 for each subsequent tier. Therefore, as the brick walls merge into tier 5 gems, we should receive
(423+33)/2 = 228; (228+4)/2=116; (116+1)/2=58; (58+3)/2=30; (30+6+423)/2=229; (229+1)/2=115; (115+2)/2=59; (59+2)/2=30; (30+1)/2=15; (15+1)/2=8; (8+1)/2=4; (4)/2=2; (2+1)/2=1.
The payouts are thus 228*1+116*2+58*4+30*8+229*14+115*23+59*32+30*41+15*50+8*59+4*68+2*77+86=11635. Substracting the hunger, we should be able to get 6163
Wow, you are playing way-way different than I was. I did fast up-down motions with the mouse, granting huge distances within short time. It filled the entire screen with crosses of all colours, which forced me to draw a slow pattern waiting while all of the spawned crosses to clear, allowing me to proceed with the pattern. It is much more risky, but as each run is vastly shorter I think it pays up in the end
Very nifty, as usual. The atmosphere is tense enough to feel obliged to solve the puzzle, yet not too scary to be distracting. I think I've skipped two artefacts: one above the big square grid, as I couldn't find how to interact with it; and one on the screen where you have to freeze a pattern - I don't seem to find the right combination. Overall, a very creative mechanic, very interesting