CGNAT indeed seems to cause unexpected issues, especially since standard fixes like port forwarding or firewall adjustments often don't help. For anyone still struggling with connection problems related to CGNAT, I found a clear explanation online that sheds more light on why this happens. It discusses how NAT traversal works, why CGNAT disrupts peer-to-peer connections, and offers some additional context that might help others better understand what's going on: NAT Traversal: How Does It Work? This aligns closely with what others have pointed out here—CGNAT introduces unpredictable behaviors in port mapping and network handling, making multiplayer connections challenging. Hope this helps clarify things further!

It has been tested and proven that having <5 very strong units is never a problem and is a fun addition to the boring cs spam. but the number is always the problem. The hero is incredibly strong as an unit but nobody complains, because it is alone. The same applied for fire Cav in A25 and spear champ now. 5 ultra op spear cav can't completely break the game but 40 just slightly above-average merc sword cav broke A25

10% is a lot actually. Caracutos (brit speed hero) is 15%. Also, speamen do not need to beat champcav in a 1:1. They currently win decisively in a 2:1 engagement (which is less spearmen than a resource-balanced fight). The issue is that a champcav player will almost never need to take a 2:1 engagement against spearmen. first you kill some vulnerable economy to buy some time and later strike with closer to a 1:1 ratio. The massive mobility gap allows that (as high as 230% between spearmen and mace champcav+hero), not so much the strength of the unit. Also, such high mobility means champ cav can always be dealing damage, while infantry must spend a lot of time simply walking. persian and seleucid champ cavs are now "cataphract" units, which are a bit more expensive, and even stronger, but slower. Has anyone else noticed their popularity slip a bit compared to gaul, mace, or roman champ cav? So I think a few options are good here: close the mobility gap a bit by removing the extremely cheap, global cav speed tech, improve the utility of palisades, walls, and buildings to constrain mobility, and if need be, bump the cav counter up a little.

Ah yes this I agree with. Sniping is quite undesirable, I'm not in defense of it I'm just against the game working entirely through buttons using optimized automatic behavior. In my personal opinion automated things are ok so long as they are outperformed by manual human control, and unit behavior falls into this category.

I don't like the tech either. But I don't think it is a big deal as the % increase is just too small. It really only matters in cav vs cav fights.  The problem is that champ cav can beat their supposed counters (spears) in straight up fights. Until that changes, champ cav will always be the best unit and optimal strategy will require players to spam as many champ cav as possible. 

A change for this alpha is that champs coming from temples and unique buildings train 25% faster.  I could bump that up a bit, maybe 30 or 35%. The issue I have with training from forts is that you had a situation where forts were just an expensive barracks that you needed a lot of space to put down. In other words, they weren't used defensively. Making them come from forts doesn't change the fact that the unit itself is strong, and it would make melee infantry champs extremely unhelpful. I think specific units could be moved to forts on an individual level, but I disagree with the formulaic approach of moving all the barracks/stable champs to the fort. IMO, this tech has to go: 

How would you do that? Just like you can't make recruiting a one-button choice, since you might want certain barracks at certain positions producing certain units and other barracks producing other units, I want some of my ranged units to attack certain enemy units and some of them to attack other units (or even buildings when they're currently under construction, to delay the building being done). I just don't see how a single button even could do that.