There's quite a bit of speculation going on here, and much of it is missing the mark.
SP3 introduced a TNI Physics plugin. This takes the previous "native" physics code and exposes it in the plugin that developers can then edit, update, replace etc. In theory, everything should work in the same fashion as it did. In practice, some edge cases result in different behaviour and bugs can also lead to different behaviour.
Portals were not "revised" at all. It's just that TNI physics was running rather than native physics. Without physics, trains don't go anywhere, so the idea of not having physics for trains coming out of portals would turn those trains into scenery objects. (In other words, portals have always relied on train physics).
Its not the same behavior though, I ran the exact same consist I tested under SP3 in the original TRS19 I have as backup and there was no derailment. I can understand that you cannot test all trains, but the tests I did was with the original "Auran" content (EF81) and they failed so its quite hard to believe that there was testing done when even the original content derails because of the physics. And that no one ran a reversed loco or multiple unit trough a portal, not realizing the abrupt stop the consists makes whenever a reversed unit is emitted? And this happens with all trains so how come this wasn't noticed?
What you said about the TNI is not really correct, the trains do emit from the portal and they do run, you just don't have control over them (Whats missing is TNIAllocPhysicsVehicle). But as soon as TRS loads TNIAllocPhysicsVehicle the trains start to jump around when coming from the portal so clearly that needs fixing.
As I said earlier the player should have a choice to choose between Native physics and TNI physics, since TNI is as far as I read still in Beta stage and especially with the variety of content.