I'd love it if someone could play through and give suggestions for balance.
I can do that.
I just finished a playthrough of 1.2.1 (started out as 1.2.0). Biters on, biter expansion on, spoilage off, nuclear power balancing on, fully default map generation settings. I built an overly large base with a main bus, fully stocked belt mall, belted science, and filter-splitter-based letter gathering. No sushi, no robot-based production, no train bases. My findings are certainly a part a product of the approach I chose to take.
Overall, I think the balance works surprisingly well. The balance is of course completely wack, but that suits the mod - making do with what tools you do and don't have at your disposal is half the fun of the mod, and some things being unexpectedly cheap or expensive because things just happen to work out that way adds to the fun and makes it authentic. I used a bunch of tools I would otherwise never consider, which is a great sign. I have never been annoyed by a particular item being too expensive (or cheap) to produce, or being available too late. All the difficulties I did write down were solvable problems in the end, and I mostly enjoyed solving them.
I did spend an awfully long time in the early game collecting the item stockpiles necessary to build a base that could build anything in a way that is reasonably consistent and reasonable speed. This is definitely partially a result of my chosen approach, and partially a consequence of me trying to level up faster than is advisable. Making a non-sushi factory in this mod that a reasonably-sized stream of what you need without clogging takes a lot of infrastructure, which took a long time for my various starter bases to produce the resources for and a long time to craft. Any balance change that reduces that dynamic would have been the most consequential change for my playthrough, but I am also well aware that "you chose the wrong strategy you dingus" is part of the answer here.
With that in mind, the following thoughts on game balance stand out to me:
Early game
Transport belts are very expensive. It takes three Ts to make a single belt piece, and you need them by the hundreds or thousands even in the early stages. This mod requires a lot of belts; a yellow belt of iron plates can support a production throughput vaguely comparable to vanilla factorio, but it takes nine belts to transport the decompiled letters to the machines that need them, so the amount of belt pieces required to do any sort of even quite early automation is far greater than in vanilla.
My late green science factory had about 10000 belt pieces in it. A sushi design is much more economic with its belt investments, but it has limitations of its own. Building up the stockpiles to make all these took a very long time, as did crafting them all. Now I definitely brought that onto myself to a large extent, but still: however you approach it, the initial investment required to automate anything in the early stages is very high. Reducing the length of the pre-automation stage of the game is the biggest possible change in terms of reducing early game tedium, I think.
One possible solution would be renaming "transport belt" to just "belt". Would that be overdoing it? I am confident it would be an improvement for players trying to build the type of base I made, but I'm less sure this remains true for other designs, and it's a not-small deviation from vanilla factorio.
Making belts produce in sets of 10 like in vanilla would work, but I think you already tried that for concrete and rejected it.
Related suggestion: move the Literacy research to somewhere in green science. Crafting the parts of a base that can automate building supplies takes a long time, and you generally want to do that somewhere before starting blue science. Literacy has its greatest value in the stage of the game before you have built automated production of automation tools, which is ideally before blue science. Which means that in its current place, it's easy to unlock literacy just after the point where you don't need it anymore.
When building the burner letter decompilers using the starter recipe in the early game (mostly for underground belts/pipes to ground), I spent a lot of time hunting for the necessary stone and wood. Chopping 500 wood worth of trees and draining 2000 stone from the starting mine (once the nearly rocks ran out) in order to make 100 undergrounds takes a long tedious time.
Suggestion: leave the fish as part of the recipe as the real limiting factor in the affordability of this recipe, but reduce the wood & stone cost. Compensate by increasing the crafting time (to the same as its decompilation cost?), and/or by making it a handcraft-only recipe, both of which will serve to make it impractical to make these things in quantity without making it too tedious when you do need them.
Production balance
When doing science, the letter I was always in short supply on is C, because it occurs twice in every science pack; I don't want to decompile too much coal because I need it for power, and copper gives quite a low amount of C compared to its byproducts that are far less important for science.
Suggestion: increase the amount of C products from letter decompiling. B, D, G, P, Q, R, S, and U could all plausibly have a chance of producing a C, which would reduce the pressure on its primary producers; if you want to be aggressive about it, you could include the I, which is far away the highest-volume letter.
The letter decompiler feels a bit heavy in its power draw to me. The power draw of a letter decompiler is the same as that of an electric furnace, which is reasonable on first glance, but you need far more decompilers than furnaces for a resource.
Cooking a stone brick in an electric furnace takes 300 kJ. Decompiling a stone brick takes 900 kJ, and decompiling the resulting letters into nothing takes about 1461 kJ. If I want to turn a double belt of stone into a belt of the better B while destroying all other letters, 12% of the energy cost goes to the electric furnace cooking the bricks, and 88% goes to decompiling the results. This seems like it should be a good deal lower.
It might be better if the default map generation settings increased the size or richness of oil fields. Compare the size of the copper ore patch you need to produce 15 C per second, with the size of the oil fields required to produce 15 M per second.
Combat
With combat and biter expansion enabled, biter evolution is a big problem. Until you have access to laser turrets, the main way to stop expansion parties is by killing nests; because gun turrets are unviable until the late game, there is no way to keep the expansion parties at bay without accumulating evolution from all the nests you need to kill. This makes for a fairly harsh timer in a mod that is pretty slow.
I think this could nicely be solved by reducing the nest-killing contribution to evolution factor. If you can defend yourself against expansions by killing nests without running out the behemoth biter clock, combat will still be difficult but more approachable.
Making military science at speed when you're in trouble is very difficult, because automating J is so difficult to scale up.
Suggestion: bump the chance of getting J from I a bit. Going from 1% to 2% or 3% would go a long way towards making military science more approachable. Perhaps even switch the 4% T and 1% J around.
Tweaks
It feels a bit strange that there is no upgrade to the letter decompiler until all the way into the postgame. The letter decompiler is the production machine you build the most of and that most determines the design of the factory, and it stays basically the same from the very first red science all the way until space science. I think getting an upgrade to it somewhere in blue or purple science would help the sense of progression. I'm not entirely sure how big of a difference a faster decompiler would make though, as I found the throughput of my belts to usually be a bigger limitation than the throughput of my decompilers. Still, it feels like there is room for improvement here.
One thing that I found frustrating is that it's not easy to keep a decompiler going at full speed when trying to recycle a belt of waste letters. When trying to put a random collection of waste letters into a decompiler, inserters generally cannot insert the next letter into the decompiler until after the decompiler finished its previous craft and the byproduct has been taken out, because the previous letter and next letter have different outputs; and factorio doesn't let the decompiler switch recipes until the results of the previous crafts are gone, which causes the decompiler to spend a substantial part of its time idle unless I carefully micromanage it.
I don't know if this can be fixed. Maybe having each letter decompilation recipe always produce 0 of each letter as a result would cause factorio to think they have the same result slots and therefore the furnace entity can switch recipes at speed without stopping? I have no idea if that works though, this problem may be out of reach.
Why don't the rail ramp and rail support recipes take a rail as an ingredient?
Overall I had a fun time with this mod. The unmitigated madness makes for a great puzzle, and the fact that certain things are hard and others are easy not because anyone designed for it but because that's just how things happened to randomly play out linguistically and yet as a player you can make it work anyway really adds something that a careful design for balance would sacrifice. The only places where I would suggest balance changes are cases where the game is needlessly tedious or needlessly frustrating.
