Since you are comparing the array to a number, JavaScript will cast it to a number. You can manually do this by doing Number(value). From my testing it seems like it just tries to cast the first value to a number. However, if there are multiple indexes, it returns NaN. And if there are no values it returns 0.

There's really not much use for the condition as far as I can imagine. It's hard to tell exactly what it's meant to do without context, but my guess is that it's a weird attempt to handle some edge case.

When you use >, the engine will first convert both sides to a primitive, which will call the valueOf, and, if that didn't return a primitive, then the toString method if it exists. For arrays, only the toString method returns a primitive, so that's what's used - and what it does is equivalent to doing .join(',').

console.log(['1', '2'].toString());

Looking at the spec again, after the array has been turned into a primitive, we now have one side that's a string (which came from the array), and another side that's a number. So, both sides are then converted to numbers:

d. Let nx be ? ToNumeric(px).
e. Let ny be ? ToNumeric(py).

And then the numbers are compared.

In the case of ['1'], you get 1 > 0, which is true.

In the case of ['1', '2'], the resulting string is '1,2', which cannot be converted into a number, so the following runs:

h. If nx or ny is NaN, return undefined.

and when undefined is returned by this algorithm, the whole > evaluates to false.

and if it would be useful in any situation?

For clean, understandable code, it generally wouldn't. Better to explicitly cast to types whose comparison makes intuitive sense first.