Did a test run on the new r38090 build, uploaded game_8256 for review. Yoral on Hard.
The AI fixes have given my enemies their balls back, so my biggest concern and reason of complaint is gone! The mid-game against Ashdar Imperials and especially the late-game against Humans featured some BIG fleets, and unlike in the previous build, I felt challenged again (or would have felt unless playing the Yoral - those damn Torpedo Destroyers are just broken). Looking at the graphs, everybody seemed to be advancing at pretty much the same pace, and overall things felt quite balanced especially in the early game. No massive AI fleets at first contact as was the case in most Hard games on the vanilla build. I think I now saw what you were trying to do to the early game balance with these changes - and I kinda liked it!
However, it still kinda worries me that smaller fleet sizes at early game give an unfair advantage to the player. The AI's are not very threatening in small numbers as they are not very smart. At equal fleet sizes, the player always wins. My two cents is that the new economy (still) leads to faster snowballing: you can go to that first big war earlier (with a smaller fleet than before) and after defeating a few ships the enemy no longer has anything to throw at you. And in mid-game and onwards, the economy changes (still) lose all relevance as you end up swimming in coin anyway.
Tl;dr: early game is easier and goes into snowball faster, but feels more "fair" and "balanced". Mid-game and onwards is not affected. If that's your design goal, I think you can go for it with this build.
...And I probably should do another test run on something else than Yoral before saying anything too definitive: those damn Torpedo Destroyers really screw up the early game balance and exacerbate the snowball effect. I brought down 9 Star Harpies with just two or three ships and nuclear-level technology. And with end-game Anti-Matter Missiles they can one-shot any capital in the game. There's not enough PD in the galaxy to stop the buggers from scrapping your ships and bombing your planets - four was enough to invade any planet with defenses. They are cheap, fast, long-range and can overwhelm enemy PD by simply throwing more missiles at them than the PD can shoot down. Don't OVER-NERF them, but I don't think removing one of those four missile slots would be an awful loss. I hate playing against the damn things, and I didn't really enjoy playing them myself either.
The Starbase spam is now gone - literally gone-gone: I only saw one Starbase in the whole game. Maybe the AI build logic still needs a little tweaking? A strategic Starbase or two might have at least slowed down my Torpedo Destroyer shenanigans. The AI seemed to fancy Space Stations instead. Shipyards were present as well.
The Arda Seed Herald tech exchange has some missing tooltip texts; I wrote down "EMP Warhead" and "Miniaturization" as having missing tooltips, but there might have been others as well. Worth looking into?
AI fleeing now seemed logical, no issues there anymore. However, it occured to me in one of the big battles that the AI could benefit from a comparison of loss percentages. E.g. if in turn two it has lost 50% of its fleet and caused 10% of casualties to the enemy, it SHOULD think "OK, this WILL end badly as I can expect this to only get worse, so lets get out of here!" Now I was able to destroy the entire Human main fleet because they didn't realize that although they had BIG ships and LOTS of them, the balance had already slided to my favour and was only going to get worse. Adding this functionality to the retreat logic could make the AI seem a lot smarter and humane, hm?
"Sparse" galaxies seem to cause some extra trouble to the AI. I fell in love with Sparse after I realized it creates buffer zones, bottlenecks and other strategically relevant borders to the galaxy. However, as the sad story of the Orthin on Bacabs shows, the AI doesn't realize when it's boxed in and won't start researching the correct technologies to get out of there. The Phidi also got a less-than-favourable starting location.
Maybe the galaxy generation parameters need to be taken a look at? Or can the AI be made to realize it's "boxed in" and do something about it? Personally I had success with building "Torpedo Raiders" (a Torpedo Destroyer with an extra-range system installed instead of a deflector screen) - that would seem like a reasonable refit in box-in situations.
Pirates and raiders are still a major issue for the AI. At least my Tinker allies did not seem very proficient at killing them off, but were wasting a lot of time and effort attacking them anyway. Every time I would have benefited from allied fleets on the war fronts, the Tinkers were instead chasing off pirates at their own backyard - and even spending something like 9 turns in-flight to get there. Maybe add a simple "do not engage pirates if in war with a major faction" logic to the AI? This would fix a myriad of issues with the current pirate/marauder balance that was collaterally hurt by the economy changes (less AI ships but equally many pirates leads to very happy pirates).
Did you do something about the routing logic of allied fleets already? Beside the pirate shenanigans, the AI empires seemed to be concentrating their forces rather well now, although the Imperial main fleet took off somewhere at the start of the war instead of continuing their successful conquests. I didn't have the scan range to see what they were doing or where they were stationed exactly. Usually alliances cause issues for AI fleet placement. They should focus on defending their own empire more instead of hanging out on shore leave in the neighbouring allied tourist worlds.