Taking it a little further: Put simply, for something to be scientific, or more precisely science-able, one needs to be able to control the variables enough to test it, yes? Anything that can't be scrutinized like that is unscientific. There's also predictability, more or less
It's helpful to think of this as a triangle.
- Subject to both scrutiny and prediction; science.
- Predictable, i.e. repeatable, but not extend-able: too many variables, or operating in a way that can't be adequately contained. Compare to inability to reliably predict weather patterns or whatever, and to a “god of the gaps” approach.
- Neither. The imperceptible whims of a god.
To achieve that sense of wonder at the fantastic while yet being able to employ it, the magic only needs operate per a set of different rules, whether bizarre or intuitive; you can expect certain things to work, but not to understand why they do so.
Something like that.