There are very many things remaining to do in historical linguistics. If you set aside certain language families which have been "mostly figured out", there are still very many areas in the world where our knowledge of language relations and especially reconstruction is, shall we say, less than ideal, especially in the New World. Plus, even in a family that is pretty well worked out like Bantu, there are unresolved details – for example, what, if any, system of vowel harmony did it have; aside from knowing what the tonal phonemes were and what roots had what tones, we don't really know what the proto-language did with those tones.
There are also broader questions that are not just about reconstruction. For example, there are various sensible ideas about how sound changes like Grimm's Law come about, but the theory of sound change is pretty coarse-grained. Being able to substantiate and generalize these ideas will take quite a long time and will require an unprecedented level of observation. An example where the horse may be out of the barn is the development of the modal + "of" construction, from modal + "have". The change is far enough along that there are now people for whom the construction is "should of gone", which we can post-hoc rationalize as phonological reduction of "should have". Did this happen simultaneously at hundreds of locations? Did it get started once and spread? How does this relate to the "shoulda" alternative (is "should of" based on "shoulda", or vice versa, or are these independent changes)?
Bacteria-model language splitting is the dominant theory of language relations, but of course nobody has ever denied the relevance of language contact, and I think that as contact-induced language changes are studies in greater detail, fundamental concepts of language relatedness are bound to change, at least in some cases (in particular, when there is significant contact, as in West Africa, but not so much in Oceanic).