Yeah ok, I see what you mean.

There is indeed as you pointed out the problem that the bandwith is shared between devices on the same electric circuit when using Ethernet Over Power but not with Point To Point Ethernet connections (which are all of them nowadays, though in the old days shared topologies were more common).

I can see how if you have devices on you network communicating with each other and using lots of bandwidth doing so, the bottleneck wouln't at all be the internet connection.

That said, point to point ethernet connections to a central router might not be the greatest solution if your running your own mini server farm from you a location away from the main router, unless you want bundles of cables along your walls. Ultimatelly keeping those things together and pulling a single cable from the main router to a secondary router that then connects point to point to all those machines, might be a better option.

And no, EoP doesn't leak in any way significantly across high inductances or transformers which have been designed to work with home electrics (i.e. for a frequency of 50/60Hz) simply because the inductors filter out the higher frequencies that are necessary to be able to transfer data at MBit/s speeds - this isn't a "networking" thing, it's Physics (and basic Electronics).

Even if the EoP protocol using is using multiple channels at multiple frequencies, the lower frequencies that might leak between your and your neighbours' electric circuits if both are using those adaptors, and could thus have too low a Signal-Noise ratio to be useable, are a tiny fraction of the total bandwith because those low frequencies which can transverse the inductors without being too dampened cannot encode and carry all that much data (it relates to the Nyquist Theorem), so you would loose a tiny fraction of the total bandwidth (I would be surprised if even a 1MHz signal can cross an Electric Meter without being attenuated to near non-existence).

Even the radio waves emitted by your wires whilst carrying that data will at worst induce tiny amounts of electric noise in your neighbours wires (and vice-versa) as neither are designed as antennas for those frequencies. We're talking sub-1% levels of electrical noise due to electromagnetic interference here.

If your house is right next to your neighbours' (such as in an appartment building), unless yours and your neighbour's electric circuits both sit behind the same Electric Meter (i.e. you share payment) and hence there is nothing but two simple circuit breakers or fuses between both electric circuits, the level of EoP interference between both will be at least one order of magnitude lower (probably much more) than the level of WiFi interference if only because of the coil in the Meter along with the natural resistance of the circuit acts as an LP filter with quite a low cut-off point thus attenuating almost all of the signal, and if you sit behind different mains transformers the signal going across will be so low as to not even be measurable.

It's not by chance that other posters were complaining that EoP did not work between devices in two different parts of the house, and that's merely them being in different electric circuits separated by simple circuit breakers.