Again, I'm not certain of the entire stack beyond my house, but the device they brought for installation was sealed in the box that said "Media Converter" (not "ONT") on the box. It could still be the case that they somehow set it up remotely, or it could be that no setup was needed. For all I know, I tested my connection with PPPoE.
They are provisioned remotely over network from controller run by ISP, which sets up appropriate timeslots (xPON is a time-division multiplexed network in practice, though DWDM - wavelength modulation - is also used).
Internet -> ISP network -> PPPoE -> OLT (head unit) -> passive fiber network (and/or DWDM) -> passive splitter close to home (often in inside appartment buildings) -> ONT -> ethernet to your router for PPPoE -> your systems
Can you please expand on how much of that chain can be qualified as "Ethernet" (or "TCP/IP") and how much is "neither at all", and how different that is from e.g. a classic "DSL with ATM" chain?
From OLT to ONT, and everything in between, is handled by non-ethernet/non-IP protocols over which you can tunnel whatever higher level protocol you want.
Technically that can continue to whatever hosts PPPoE gateway. Ultimately, OLT is "root node", and ONTs are "leaf nodes" of a passive optical network tree. What is distributed over it is less of interest to it. But that's why you often have ONTs called "media converters" - though they aren't exactly that. A typical ONT also includes xPON MAC component as well as all management functions necessary to login into xPON network, establish session, etc. and demux transported protocol to whatever is delivered on the other side (nearly always ethernet)
