Does English always obey the maximal onset principle?
Not obviously, but one can concoct analyses that dispose of counterexamples. The consonant [ŋ] does not appear in a syllable onset, thus orangutan has /ŋ/ in the coda, not the onset, unless you decide that because of the maximum onset principle it must be in the onset as well, in which case you could posit syllable-initial ŋ. If you work through the facts in Daniel Kahn's dissertation, you will see that there is no compelling case where C in the environment V_V must is exclusively in the coda, it can always be shared between the preceding coda and following onset. In which case, the discussion turns to whether max-onset is an absolute linguistic universal that can never be overridden by language-specific facts (therefore it defines how you have to analyze the facts), or is it a scientific hypothesis that might be entertained in a language, but might not be if there isn't good evidence for it in the language.
Incidentally, the name "maximum onset principle" refers to a class of related ideas, one being "onset first", the other being "everything possible goes into the onset". Kahn's rule was to put as much possible into the onset but this was just a fact of English grammar and not a "principle" in any sense. Clements & Keyser on the other hand had a "principle" of first building a CV core syllable, then adding more things to the onset and coda, which is another view of "onset maximization". So you have to also say exactly what you mean by "maximal onset principle".