RE: SO
Spoiler
I think it is very powerful for the player flagship, purely for the large offensive boost letting you punch up a lot in the early game. But it relies a lot on the players ability to engage effectively because of the large range disadvantage. Particularly identifying moments when you can get into range without taking too much fire, and also in dodging shots by strafing or shield flickering to engage without incurring a big flux disadvantage. This is greatly exaggerated in a fleet context vs in a 1v1/sim context as there tend to be many more opportunities to engage/commit.
I think the factors above really clash with the AI. This isn't purely a SO thing either, the AI is just generally not very good at identifying openings and hard committing when at a big range disadvantage, and won't pull tricks like shield flickering/dodging on approach anywhere near as consistently as the player. You can overcome some of this by trying to micro ships with orders (basically play like an RTS), but the game isn't really designed for that to be an enjoyable experience, or even really a viable play style IMO.
I don't think it's necessarily bad for the AI to be somewhat non-commital unless ordered, and I also don't really think it's a flaw of the AI to not replicate all the tricks that the player can pull, but it does mean that SO is not nearly as good for the AI, unless you are in a situation where you can just hit 'full assault' due to numerical superiority. The AI is not going to eke out a disadvantaged situation using SO.
Overall, I just don't like it because I think it's boring game design. Basically, it makes a lot situations pretty binary, where you either can rush things and obliterate them easily, unless the timer runs out and you lose, or you can't even get into range and can't get anything done (whereas a normal build could hang on the edge of range and chip away/contribute, even when outgunned).
Also, regarding end-game viability, it is obviously not going to be viable because of all the very harsh restrictions aimed specifically at end game. Particularly the PPT penalty and the restriction from capital ship, and also that it cannot be built in. IMO, the number of cases where the only balance solution was to just arbitrarily prevent interactions (or effectively do so with the PPT limit), indicates that it's just too many drastic stat alterations (speed, range, flux) to be effectively balanced.
RE: Odyssey
Spoiler
I think it's once again a very strong player ship that isn't as great in AI hands. Actually for a lot of similar reasons as SO. It usually has a range disadvantage against ballistic weapons (700 range large energy weapons), unless you use beams which just make it under-gunned in general. It pretty much needs to use plasma cannons to get the level of sustained firepower necessary to justify its DP cost in a large fleet battle IMO, although maybe s-mod expanded mags auto pulse or VPD can compete now? I haven't actually experimented much with that. It also has tons of small energy slots that are pretty much just bait IMO since they just exaggerate the range disadvantage. IMO, it should use fighters for PD/anti-fighter and pretty much leave all the small energies empty, except maybe some burst PD around the engines, and maaaaybe a couple IR pulse lasers if you have some spare dissipation.
The player can use its burst mobility to GREAT effect to overcome the range disadvantage, but the AI will not do that. Particularly the AI will never do things like using the mobility laterally to dodge shots, or using it to disengage away by turning around. The player is also generally much better than the AI at judging when to commit with mobility to finish something, or chase down small ships. For the player, I think the plasma burn is a much more powerful system than plasma jets, but unfortunately, the AI doesn't really understand how to use it.
Also, the AI will foolishly deploy its fighters across the map which both disables the zero flux speed boost, and throws away its best source of PD. Defensive targeting array does solve this, but it's kind of an OP tax to get the behavior the player can get for free, unless you use defensive fighters like xyphos and sarissa. I like to use a spark/xyphos combo on my flagship for what it's worth although wasps can be a spark alternative, and sarissa can also be good.
Some other thoughts:
Spoiler
I think balancing high skill mechanics/ships for the player will naturally lead to them being underpowered for the AI. The only way around this is to either make the AI more similar to the player, or let the player have some OP ships so that they are all balanced in AI hands.
Also, I think the AI's struggles with short range high mobility ships are somewhat universal, although most clearly seen with SO. It is quite apparent when fighting big fleet battles with small fast ships, particularly when outnumbered. They will spend most of the time just floating around doing nothing, because they cannot find a place where they feel they can safely engage. I really love using scarabs (it's unhealthy really), and have seen this issue a ton. It also pops up with AI auroras a lot IMO. These ships also have issues with using their mobility systems to engage, and then ending up in a really dangerous spot, or using their mobility system to do nothing except float around out of range.
Often, I find I need to either micro manage really heavily, or do maneuvering of my fleet to try and get my fast short range ships to the enemy backline so they can harass, pick off reinforcements, and generally get into more local battles instead of big battle lines. If I let scarabs do whatever they want, they will just hang around the big battle line, never contributing because they can't safely get in range.
Maybe it is a worthwhile avenue of development to try and improve AI in the area of Macro positioning (i.e. moving around the battlefield to find a more advantageous situation).
That can be defensive (recognizing danger before flux levels rise and running towards allies), or offensive (trying to recognize when it is not making progress/contributing, and have it more aggressively flank to try and contribute to the fight in some way).
The calculations there get complicated though, because you are predicting rather than reacting, which is inherently much more complicated.