Kripke (Naming and Necessity) described how "water" means H2O because people once indicated the liquid form of H2O and said something like "that stuff," so that (via a causal chain) we now mean by "water" what they meant by it. But if so, then had some weird spacetime wormhole (or something) substituted that bit of water for XYZ briefly, just before they said "that stuff," then we would have been misusing "water" ever since (and obviously we have not, whereas it's not so obvious that such a substitution could not have been).
......Furthermore their folk metaphysics may not have been much like ours, way back then (instead of our chemistry, something like alchemy maybe), so why would "that stuff" have meant H2O then? Indeed, what stops us referring, with our "water," to H2O in any of its fluid phases (the liquid phase is largely H+ and OH- ions anyway), or to any similar mixture of hydrogen and oxygen (similarly) or of nucleons and electrons, or to just the oxygen or just the nucleons, or (conversely) to impurities in the water as well as the H2O (as we may well do ordinarily), and so on?
......Presumably we all presume some (similar) folky metaphysics, so that our "that"s are intimately (if subconsciously) associated with something like a description (if an indescribable one) in our heads. Furthermore, once we've separated out such descriptivistic content, there might not be anything externalistic left over.
......Even if the reference of "water" is fixed from day to day, by our thinking "that stuff" while thinking of some actual water, the substance underlying H2O might be a different one on different days (e.g. as subatomic strings randomly pointed in different directions, or something), and we would not then mean, by "water," different substances on different days. If the underlying variation made no observable difference, it would be like there was (as presumably there is) no variation; the stuff referred to would be the constant chemical (defined by its chemical description, and hence via the meaning of "chemical"). And if it did make a difference then descriptivistic content would again determine which stuff we were referring to (cf. normal versus heavy water).
......Furthermore, while there may seem to be an externalistic element with the "that" of "that stuff," there would surely be (after all the descriptivisitic stuff) an internalistic "this," because only a reference to this Actuality (containing both that stuff and us) would be able to lack descriptivistic content.
......There is therefore (for a frivolous consequence) surprisingly little incoherence between Biblical literalism and modern science; e.g. when originally creating water (not just below but also above the starry firmament) God would have had some description (some range of relational roles or whatever) in mind, but such details as the underlying substances (e.g. the particles that comprise the H and the O nowadays) might quite naturally have varied as the more crucial (e.g. moral and psychological) matters were settled.