Yup, you guys are right. I completely failed to realize or forgot that diagonals count for adjacency as well. It should push the expected per tile yields for a realistic energy farm up to ~6 from ~5.4, which with a quick back of the envelope should mean that it'll be very close FOP wise pre-fusion, though still dominated post. I'll work up more detailed numbers later.
edit: Okay, so a 5x5 energy farm starting with the Sunny Mesa costs 2027 credits, or 81 per tile, producing 152 energy/turn. The cost goes down to 1387 if you can surround the entire thing with bases before starting, though I can't imagine anyone would ever be able to do that. Since a 5x5 is about as big as you can plausibly afford on standard sized maps, this should make a good baseline. Since there's no point in comparing with specialists thanks to the cap, I'll only compare land vs sea.
Land costs 312 terraforming turns for raising, and 180 for the improvements, plus 750 minerals of crawlers and 2027/1387 energy. You need 38 shelf tiles to match that, needing 152 terraforming turns, 1900 minerals worth of trawlers, and a 80 mineral facility. Canceling out, land vs sea yields 340 terraforming turns + 2027/1387 energy vs. 1230 minerals.
Post fusion, trawler cost goes down to 1140, changing it to 340 terraforming turns + 2027/1387 energy vs. 470 minerals.
I did have another thought though, where you only raise a 3x3 square in the center of the 5x5 farm. The outer ring would be raised to 2000m+ without any expenditure from you either in terraforming turns or cash. Efficiency declines of course, with output falling to 136e/turn, but costs fall to 120 terraformer turns and 756/628 energy for raising land. You'd only need 34 shelf tiles to match that, so prefusion, the comparison goes to 164 terraforming turns and 756/628 energy vs 1030 minerals, clearly favoring land. Post fusion, of course that last goes down to 350 minerals. Hell, you can go even further and only raise the center of the Sunny Mesa (amazingly, this seems to work for just as well for any random tile over 1000m) for just 24 terraforming turns and 125 credits and get a 5x5 with a 3000m center, surrounded by a ring of 2000m and then with a ring of 1000m, producing 112 e/turn when fully developed, driving the prefusion comparison to 92 terraforming turns and 125 credits vs 730 minerals, with the latter going down to 170 post fusion.
This is all up to alternate interpretation of course, but my conclusion is that the main determinants would be 1) how valuable land is (deriving from map size and how close enemies spawn) , 2) just how long a can land energy farm plausibly or beneficially be fully set up barring finishing raising the tiles before Fusion, and 3) whether or not your maximal multiplier base was going to build a thermocline transducer anyway. If (2) is "not long" and (3) is "yes", sea is clearly dominant. Otherwise, it gets a lot trickier.
.....
