Banning opposing directions each time the stock controller has happened to ban them is not a consistent rule.
As I said, we don't TAS arcade cabinets, or physical controllers, we TAS the software in isolation. If we throw physical device factors in, we're not emulating anymore, we're simulating. And perfect simulation is impossible, let alone non-deterministic. Even if we start this simulation of a physical device, where shall we stop? Let's simulate temperature as well, which will allow us to bend the plastic enough to be able to reach those opposing directions physically. Then why won't we also simulate all the non-determinism caused by different initial RAM on different consoles? Why won't we simulate half-connected cartridge and its glitches? Why won't we simulate kicking the console while holding a dog that has just bitten a cat that was sleeping? I want all the factors to be considered, what the hell?
Then, we can't base on the lack of information. Maybe someone has just done poor research and because of that they state that no official controller allowed opposing directions. Then someone else does better research and finds the one that allowed that. Like
https://en.wikipedia.org/wiki/Power_Pad Now if someone declares that only the controller that went with the console itself is allowed, like SDA do, then we hit the first problem I described.
Finally, this technique has been there forever, it won't be possible to ban it anymore.
However, it can happen that opposing directions get banned by the TAS author as a way to improve entertainment and provide a low-gitch run. If it meets the Moons criteria, it
will be published.