I always find these threads interesting to read. Just as in reality though, sometimes I think one needs to separate the physics from the engineering.
In this case, there is a Trek physics problem that needs to be solved: Bending, warping, subspace bubble, whatever it is that moves the ship forward.
Then there is the engineering to solve that problem.
Now, reality suggests that any given problem has multiple solutions, even multiple optimum solutions, so I see no reason for that not to be the case. Just because coils are required for one solution doesn't mean they are required for every solution.
Reality also suggests that engineering solutions are tailored for their implementation. While two designs share certain general physics similarities (steam turbines for example), one cannot always infer what is true in one is true in another (turbine connected to prop shaft vs turbine connected to generator that powers
azipods.) Just because the dilithium crystals are integrated in one design does not follow that a second design integrates them as well. You don't necessarily divorce TOS from the Trekverse, but you can divorce the Connie from the Galaxy. (Variations occur even within ship classes so you could even divorce Defiant from Enterprise LOL. (A
Flight IIA Connie?))
Reality also suggests that engineering systems are rarely static. They tend to be in a cycle of use, upgrade and replace. So maybe there wasn't a M/AM reactor in the hull at first but an upgrade puts one there later.
Point of all this is that it is okay for the Connies to not have a warp core, not have coils, etc. It doesn't break continuity and, since such variation mirrors reality, it actually adds a dimension of realism to the mix.