is that you're not sure if its (mostly) on or (mostly) off, or which way to turn it if its not!
Consider this - you hit the water, suck, and get nothing.
You reach back for your valve. The first way you turn it moves. You continue and hit a stop. You just CLOSED IT, and still have no gas! Ditto if you have it on and need it off NOW.
If its against the "on" stop, but just lightly, then you KNOW which way to crank it for the opposite state.
Now on a SHIP, yes, FredT is right. Same for process control applications. But Scuba is neither, as valves in the former two are often not moved for days, months or even years until suddenly they NEED TO BE moved RIGHT NOW.
Scuba valves today are a bit different. When you need to move them, you need to move them RIGHT NOW, but you also need to know with some reliability WHICH WAY to move them, and you're usually, when in that situation, not looking at them at the time.
If, for example, you lose a burst disc underwater on a manifold, you must isolate quickly, or you risk losing ALL the gas in both tanks. Turning the isolator the wrong way could easily cost you enough gas to screw you. Since the "usual protocol" is to close the post you believe is leaking first (since the most common failure is in the first stage, not the burst valve) if THAT is also not up against a stop then you potentially have to make TWO mistakes before you stop the leak. That's quite a bit of extra time in a situation where you don't have any.
I want my tank valves LIGHTLY against the open stop when in actual use. Not hard against, just lightly against. If I try to turn it the wrong way in the heat of the moment, I thus know INSTANTLY that I've done so.
If the valves are stiff or otherwise hard to turn then you need to service them and fix that. Unlike gate valves and such which may not be torn down for years at a time scuba valves usually either are or should be.