Some additional observations about Erebus map generation and game balance in general:
The subdue animal promotion seems incredibly strong. I find myself always going animal husbandry first. It is somewhat of a gamble though. Maybe lock it behind hunting too (which really needs a boost imho)?
Floodplain starts seem very strong. There isn't nearly as much drawback to them as there was in the base game. Forges are hard to get and produce very little unhealthiness. Health is very easy to get, I never find myself limited by it.
Desert also has the best resources. Gold and Incense are top tier luxuries. Stone and marble are strong.
The AI consistently does much, much better in floodplains then anywhere else. A decent floodplain start is ~2 whole difficulty levels easier then playing against floodplain started AI.
The only AI that does really well without floodplains are Lanun, their pirate coves are just that good.
Where floodplains are really good, jungles are prone to mess up starts. The Lanun seem to be the only ones able to cope. In normal civ, iron working is a staple and essential tech many strategies want to rush. Advanced civ treats jungles as barriers for start positions, much like a desert and never places civs in the middle of them. Jungles being poor start positions is especially bad for civs that have other priorities then melee.
The Balseraphs tend to perform poorly in my testing, but I assume their jungle starts in my tests bring a lot of bias.
Illeans and Dovello are almost always last, because tundra does not provide strong enough resources to balance having to start there. This ties into the idea of locking subdue animal behind hunting: To buff tundra starts. They may need more then that though.
Dwarves are often doing very well. But that seems mostly because their starting position bias tends to place them near floodplains.
Falamar/Lanun doesn't care as much about his starting position, pirate coves are very strong.
The Ljosalfar are very volatile in their performance. If they get ancient forest clusters or floodplains, they go brr. If not, they often place second to last (before the tundra start races). The main reason for this seem to be poor AI weights though. Their workers do not prioritize improving forest plots. They tend to get calendar+agrarianism even without resources to improve, when they really should have prioritized cottages and Way of Leaves. The proposed -10% workspeed probably makes a bigger difference on fast game speed, because 1 turn is more valuable there. I don‘t think worker speed is the main issue for the AI Ljosalfar though.
Cassiel/Grigori just never does well. I have too little understanding of how Grigori work to give useful advice.
The Elohim tend to do below average, except for floodplains starts.
Some suggestions for changes if you are still willing to implement changes:
1. Make the subdue animal promotion require hunting (maybe in addition to animal husbandry?). This would buff tundra locations because they naturally want hunting for improvements and they usually have a lot of space for animals to spawn in.
2. Reduce available health, increase unhealthiness from buildings and/or population. Health is just too abundant, floodplains have no real drawback. Reducing the amount of health granted by granary and smoke house would probably be a good start. I‘m not sure I like the granary change compared to vanilla. Granaries are supposedly not OP in vanilla except with slavery.
3. Amurite balance: Replace fire 1 with something useful for combat. Maybe a firebolt that is a smaller fireball with range 1, 2 fire combat and doesn't deal collateral damage. The idea is to improve adepts in combat for Dain/Amurites without any other tech.
4. Lock dancing bear, cobra charmer and the griffon aerie behind carnival (or some other building), like the rest of the animal pens. It seems a little inconsistent that some of the happiness animal buildings require a carnival while others do not. It also makes those much better then the animals requiring carnival.
Halving bonuses from animal buildings (round up) seems like a good idea too, the bonus is too large for the amount of rng involved.
5. I can't seem to make supplies after constructing a bathhouse. Is that intentional?
6. Change the Ljosalfar AI in the following ways: They can probably rex a little less and play taller, even though two of their leaders suggest otherwise. Heavily de-prioritize Agrarianism. Calendar should be low priority, like Mining: Only get if you need happiness and/or have lots of tiles to improve. Prioritize Education and cottages more. Workers should prioritize improving tiles with forests first. Change Arendel's favorite civic to something sensible like guardian of nature.
7. Illians balance: Maybe give their temple of the hand a commerce bonus? Maybe just a fixed commerce on the building. Their lack of commerce is probably what makes them so weak. You could also discount the temple with stone, I think it would be thematic and play to their start position.
8. Ljosalfar balance: I'd probably change ancient forest placement before touching Ljosalfar aside from making the AI play them more sensibly.
9. Lanun balance: The cove nerf already proposed seems warranted.
10. Jungle is worse then tundra in my book. I think start position placement should consider this more (advanced civ does a god job here, maybe those changes can be transported?). Due to the different tech structure compared to vanilla, I think remove jungle would be better placed in a tier 2 tech.
11. Elohim could use the old defender trait as 3rd trait imho. Dunno if trait slots are limited by number though (it seems like sage replaced it?).
About the map script:
Advanced civ has some really good solutions for civ placement. Maybe those can be ported?
Advanced civ also reduces clustered gold. Gold is very strong, reduced clustering would improve map balance imho.
Maybe there is a way to balance placement of special features? Some are super strong (remnants) while others can break a start position (guardians). Generally speaking, I think special features are more interesting then raw mana nodes, distributing them fairly and more frequently seems desirable.
I think it would be better if ancient forests would be sprinkled around the map in single tiles only, not clustered (if enough regular forests are around). Clustered ancient forests come close to floodplains in start position power but don't quite seem to match them. Maybe place some in the tundra too, tundra starts can use some help. This should also help the elves with their economy problems.
There will probably be more, but I really need to go to bed now. I spent way more time with this then I should have
Thanks you for maintaining this great mod!
Edited multiple times for better wording and more ideas.
) called Vinlandsaga where the players started clustered on a tiny island and had to break out onto a mainland inhabited by skraelings. I quite enjoyed it. I can see being frustrated by clustering, but I think I'll be trying Erebus/Erebus Continents next 
