There are very significant consequences to epistemology, though, even if the being in question is completely non-interfering, undetectable and inconsequential (if such a being then can be said to exist, however, is another question altogether). And science depends crucially on its epistemological groundwork – without it, the whole edifice, and along with it our ability to make reliable statements about reality at all, comes tumbling down. It’s not so much an issue whether or not we’re forming hypotheses about the world right now; the issue is that the ‘maybe’ stance you’re arguing from makes it impossible to form any hypotheses at all, ever, and expect them to meaningfully apply to the world through any means other than blind chance.
I think the fundamental thing to realise is perhaps that there always exist innumerable distinct hypotheses that explain any given dataset in accordance with all available evidence. What your stance now amounts to is that since all these hypotheses are equal regarding their explanatory power, all of them are equally reasonable. This is what you’re saying when you claim that if the existence of a being is not decidable through evidence, you can equally well believe or disbelieve its existence, just phrased somewhat differently.
Now, most (almost all) of these hypotheses generally will differ regarding the predictions they make – for example, the fairy-based theory of gravity might predict that there’s no gravity on Sundays, since fairies get the day off. So, well, that’s easy, then – predictions are testable, and we just throw out all the hypotheses whose predictions didn’t turn out right. This sounds swell at first, until you realize that even then, you still have innumerable possible and equally well evidenced hypotheses after each prediction you test. For instance, while fairy-gravity might be disproved by throwing rocks on Sundays, fairy/elf gravity, in which the elves take over the fairies’ duties on Sundays, isn’t. The consequence of this, then, is that at any given point, we have innumerable different hypotheses, all making different predictions – we couldn’t dare to try such things as, for instance, trying to send a rocket to the moon on a Sunday, since the moon-elves are at war with the Sunday-gravity-elves, and the rocket wouldn’t hit its intended target due to sudden gravitation failure. Or anything: we could simply not make predictions about the world with any confidence.
But, you’ve stipulated a non-interacting god, who wouldn’t do such silly things as having gravity just fail all of a sudden. So it would seem we’re safe then, since instead, he always acts in perfect accord with… what, exactly? The laws of nature? Scientific theory? Well, of all those innumerable hypotheses, which one would that be? How do you now decide whether or not a hypothesis is actually ‘scientific theory + a non-interacting god’ (or, if you will, 'the way the world would work if there were no god + a non-interacting god), if there’s no way to know which one is the ‘scientific theory’? This hypothesis would, along with all of the others, just swim around in this infinite pool of indistinguishable hypotheses, with no way for us to pick it out. So, if you wish to admit the ‘scientific theory + non-interacting god’ hypothesis, you have to admit, on the same footing, all of the others – fairy-elf gravity, the world being ruled by shape-shifting octopodes from Alpha Centauri, Lee Harvey Oswald being controlled by a secret cabal of invisible dwarves that lived in his brain, etc.; all hypotheses that conform to the evidence, and since every hypothesis can be made to conform to all possible evidence through the assumption of enough ad hoc elements, that is basically every hypothesis you can think of, and every other.
Thus, the only way to form reliable hypotheses at all is to form them on the basis of exclusively assuming that which is necessary to explain the evidence. This doesn’t mean that ‘scientific theory + non-interacting god’ is necessarily false, or even that there can’t be an interacting god (for it would be trivial for any omnipotent being to hide itself from us), or that fairy-elf gravity is wrong; and much less that you should unquestioningly believe in the truth of even the most parsimonious hypothesis (all hypotheses are always only provisional, after all), but it does mean that you can’t assume parsimony-violating hypotheses a priori and expect to be able to form reasonable statements about the world. The default hypothesis has to be the one lacking belief in any entity not necessitated by evidence; you can either form your hypotheses according to this maxim, or you can’t expect any hypothesis you come up with to bear any relation to reality (other than through pure chance). This is a rigorous dichotomy.
I hope I’ve made clear that the question whether or not an entity tampers is entirely beside the point, that the distinction between parsimonious and non-parsimonious hypotheses is an in-principle one.
Everything one can talk about has bearing on the real world, if only by virtue of being something one can talk about.