Also, assisted pad steering - in my opinion - doesn't belong in a discussion about driving physics since FFB is such a large part of it.
FFB is it's own separate discussion. You can turn the FFB off on your wheel and it doesn't change the physics at all. It changes the driving
experience and it changes your ability to most effectively interact with the physics in a dynamic situation, but it's not at all necessary in order to evaluate whether the fundamental physics are behaving correctly or not.
Likewise, steering assists and the like don't change the fundamental physics either. They can make it harder to figure out what's going on if the assists are not clear about what they're doing, but it's ultimately possible to evaluate the physics even with assists. The exception being assists like SRF that actually do fundamentally change either how the physics operate or key variables within them.
What you seem to think of as "driving physics" is a combination of the actual physics system of the game combined with the input and output/control systems - basically the whole chain of subsystems that a driver is going to be interacting with. If you're primarily interested in how the game races, then I understand that perfectly. It's a practical approach.
But there's been a lot of games out there in the last decade or so with really quite good physics but shockingly abysmal controls. Stuff as simple as a shoddy brake mapping curve can make a game very difficult to drive. Reversed FFB makes a game damn near undriveable. Because "hardcore" sim games are so niche a lot of them seem to release in early access or something similar, and so it becomes worth talking about the underlying physics separate from the control systems. Control systems are often easy to fix, either by the developer themselves, the user tweaking options or the mod community creating patches. Physics rarely change in fundamental ways.
That's why you'll see people like myself who like talking about the physics systems from multiple games usually talk about the physics completely separate from as many other functions as possible. And why I say above that control method doesn't matter for talking about the physics. The point is to evaluate the physics themselves, not the physics + the controls.
Good underlying physics is important, because the best FFB wheel in the world doesn't help if the car physics are wonky. But a game with good physics can still be raced clean and fast on a pad or a joystick or even a keyboard, it just takes the developer doing the work to overcome as many of the limitations of the input device as possible.
A good FFB wheel is always going to be best because it best replicates the real life controls of a car. But a pad is fine, and if people are bad drivers with a pad it's not the pad's fault, it's because they're mouthbreathers who think that the fastest way through a corner is by bouncing off the side of your car. People can be **** with wheels too, it's just less common because people that spend hundreds or thousands of dollars on a peripheral have probably taken the time to get at least moderately decent with them.
When you say "most pad players are erratic" it's actually just a combination of "most players use pads" and "most players are absolute garbage at racing". It's not an innate pad problem.