The 1d6 damage is granted as a class feature. Normally, when a natural attack is granted via a class feature, template, or soulmeld, the creature's default size is assumed to be medium. Sometimes whatever grants the natural attack tells you to adjust up/down for size differences, but if it doesn't, then the size adjustment is generally implied.
If I'm a medium-sized character that takes a level in Renegade Mastermaker, I get a battlefist that does 1d6 damage by RAW. But you just said it goes up to 1d8 because I'm one size larger than small. (Which you said is so it matches the ECS battlefist... but you declared they were two different things, so why exactly does the damage have to match if they're two different things?)
Sorry, I was just clarifying. No, you didn't say otherwise. Normally, light weapons can't be used with PA. This battlefist is an exception, because it counts as both a natural weapon and a manufactured weapon.
Again, I was just trying to clarify. Normally, a slam attack can't be used as an offhand attack because it's a natural weapon, and those are either treated as primary attacks or secondary attacks by RAW. The text doesn't specify if the slam isn't your primary, do you treat it as a secondary attack, or can it be used as an offhand attack? By my reading of the rules, it can be either, and I just wanted to clarify that.
I have two arms. If I replace one with a Battlefist, I still have another arm. You said the two battlefists were different, so I could mount one as a class feature, then get the mighty arms graft (which is a graft, not a magic item, hence it doesn't take up a "slot"), and then mount the ECS battlefist on my other arm. (The battlefists are enchanted as weapons, not as arm-slot items, so I'm not sure why magic item slots would be an issue.) If this is possible, then the ECS battlefist bumps up the slam damage on my RM battlefist by a size category... and creates some perplexities if I'm considered a Warforged with monk levels.
By RAW, adding the Living Construct subtype to a humanoid doesn't change the type to Construct. It's a subtype, not a type. In this case, there would be no Schrodinger problem: you're still a humanoid, so you still qualify for the PrC.
I was arguing that RAI makes more sense... the designer intended the type to change to Construct, with the Living Construct subtype. In this case, I'm perfectly happy with your "No Schrodinger" ruling.