Thought I would take a shot at answering your questions in a straight forward manner.
Unfortunately that means I have to gloss over quite a bit. For the non-glossed over version, read through all the posts above
2 NEOs give you more available power than 2CIMs. It could be described as “close” if you are comparing it to 2 MiniCIMs or something.
You are right, that is excessive. You are able to get more power out of 3 NEOs than 3 CIMs, but it probably isn’t making that big of a difference on the field. IFF you put 16 NEOs on each side of your drivetrain, you would be pulling more current than the FRC battery could provide and you would not be getting any benefit. Your robot might not even move! So there is a balance, but 2 NEOs per side is a pretty darn great amount.
Personally, if I want a high level comparison between motors, I look at their Peak Power rating, AND the power they deliver at 40A (like you posted). Are those two numbers the whole story? Definitely not, motors are dynamic and spin differently based on “where” they are on their operating curve, but it gives you a useful metric. Also keep in mind that while each motor is limited by a 40A (max) fuse, those motors can still pull greater than 40A for short periods. The fuses won’t trip until they are over 40A for a bit (you can look up the exact timings) so even the power at 40A number isn’t perfectly accurate.
Not really, no. You can hit peak power of some motors if the motor will only draw current a bit above the 40A limit (60A is probably fine), but the average team probably wouldn’t notice it. You may be able to see better acceleration, etc.
It’s more than enough. For the past couple of years teams have been showing that 2 CIM drivetrains perform well enough. 254 and many other high power teams have been running that. I would guess that they will be switching to brushless motors though. Not so much for the extra power (though that is nice) but for the improved efficiency, lower weight, and improved packaging.
Depends on how big your reduction is. What will really matter is how many pulses you get per revolution of your output shaft. But to your point, putting the encoder on the output shaft is usually better anyway.