> everything that is obsersed rather than conjectured about the universe implies that it has no "outside"
"outside" is not very rigorous. For example, we seem to be observing objects vanishing behind the comoving horizon. It seems unlikely that the individual cosmic horizons centred on each of our microscopic components is destroying these objects, and it seems unlikely that they will ever reenter the comoving horizon. The metric expansion induces other interesting horizons, too, and each has an "inside" and an "outside". But how many of these horizon-crossings are directly detectable by the objects crossing them? (Reflectively, we are each exiting the horizons of distant observers at slightly different scale factors because we aren't occupying the same point in spacetime).
"More of same" for some distance outside of e.g. the cosmological event horizon is wholly reasonable. We can even put lower bounds on "some distance" depending on how we look at the homogeneity and flatness problems. They're big. IIRC Guth's original cosmic inflation work predicted that the Hubble volume is no more than 10^-26 of the total causually connected volume at the start of inflation. We can also put bounds on any sort of gradient on various apparent constants such as c and G, and the region in which they are virtually certian to have the same values we have experimentally here is also big.
However we currently can't do much better than that. The bits and pieces that were close to us in the hot dense phase of the universe are mostly inaccessible to us now, but there is every probability that they have identical local physics to us. There may be bits and pieces that were insufficiently close to us in that phase that have wildly different physics; and we do not really know anything about the still denser phase of the universe. It is perfectly reasonable to consider observables generated by detailed guesses that e.g. avoid an actual singularity like Carroll & Chen (who propose one or more other universes evolving towards de Sitter space from some arbitrary shared values surface) for instance. But even there, "outside" gets tricky.
Moreover, "inside" vs "outside" is not really the best way to approach the underlying question, "what is the nature of the metric expansion of space?" where the real answer should identify the cosmological frame and its preferred coordinate system in which almost all matter remains at essentially the same spatial coordinates from the big bang to the infinite future, with the scale factor relating to radar distances (and radar beam wavelength changes) between objects at distant spatial coordinates. Then we can admit that there are various ways to interpret the radar observables, with the "easiest" one being a purely local evolution of the vacuum at each point along the radar's path, which in the preferred cosmological frame can be seen as dark energy, but which with other systems of coordinates can be seen as anything from "the local creation of more space" to a Doppler effect to (somewhat less usefully) a change in the vacuum's refractive index.
> what is North of the North pole
Well, a change of coordinates on the Earth gets rid of that particular coordinate singularity, doesn't it? We can get rid of all sorts of oddities by swapping the coordinates we're using, which is probably the greatest strength of general covariance.
What we can't do by changing coordinates (even to accelerated systems of coordinates) is eliminate the Earth's oblateness.
Likewise, we can change from FLRW coordinates to other coordinates on the cosmological frame and get rid of (or introduce) all sorts of oddities. We can't, however, get rid of angle-brightness-redshift relations.
It is a bit silly to introduce all sorts of additional action-at-a-distance forces to explain the Earth's oblateness in an effort to do away with the rotation-oblateness relation for objects close to hydrostatic equilibrium.
Similarly, it could be a bit silly to introduce all sorts of new (and presently broken) symmetries in an effort to do away with away with causally disconnected regions (with possibly different physics) before inflation or whatever mechanisms produced the relic field anisotropies. And I think it's reasonable for physical cosmologists to argue that it's hard to foreclose on that type of multiverse without doing so.
Moreover, "before the big bang" is perfectly reasonable if one starts adding in new symmetries anyway; if quantum gravitation avoids the BB singularity (as practically everyone hopes it will avoid BH singularities), why wouldn't it be meaningful to consider the universe before the classical singularity BB? It's just technically difficult as a theoretical programme and is unlikely to be guided by current data. Well, so what? It's technically difficult to predict early structures or them right after reionization, let alone say anything about or observe anything at slightly higher redshifts, and we don't even expect new physics to be relevant in those regions at all.