Well,
Species are for the most part artificial constructs, usually based on the opinion of a taxonomist. Don't confuse taxonomy with nomenclature - taxonomy is a classification - a system, whereas nomenclature encompasses the study of validly and invalidly published names used in classifications. Non-objective classifications involve 'hand waving'. That is, taxonomists working purely at the alpha level who do not try to subject characteristics (e.g., morphological, molecular, karyotypic) to objective phylogenetic reconstruction make non-objective taxonomies. Of course, choice of characteristics can be non-objective as well, but minimizing unnecessary assumptions (Ockham's razor) will always be more powerful than utilizing more.
So, where does this leave one with lumping vs. splitting? Well, I'd first want to make sure I'm using the most objectively derived classification. I'll put in a vote for *basically* phylogenetic ones based on a particular body of data and a repeatable algorithmic procedure (did some of this myself). Then I'd make sure that I was following the rules of nomenclature for proper use of names in my classification. Of course, one must also bear in mind that a particular collection of data may not be entirely representative of the names chosen to represent that data! As has been pointed out, population-level variation can be significant, and the actual status of all slipper orchid gene flow will likely not be determined in reasonable time.
So, a classification winds up with a taxonomist encompassing a certain amount of variation into what he/she calls a species. That amount of variation is non-objective, and will not be revealed by standard phylogenetic methods. No use even thinking about it. So ----- names on plants, in my opinion, should follow as an objective look at evolutionary history as possible, then take the species level with a grain of salt where variation is great. Paph. armeniacum and micranthum, e.g., pose no problems -- but lots of Lorifolia Phrags do, hence the difference in what taxonomists argue about.
Nomenclaturalists will go about the task of validating names, describing new taxa, etc., and this is basically bookkeeping ---- not biology, and the endeavor does NOT result in better species concepts. Remember the difference between taxonomy (systems) and nomenclature (assigning names to be used in systems). The best species concepts, in my opinion, are (again) those that require the fewest number of excess assumptions to maintain them. So, one may need to lump sometimes in order to retain economy of assumptions, and split other times -- all based on whatever evidence is at hand -- and I believe that evidence should be made available in a form such that other persons can repeatably arrive at conclusions as often as possible. I co-described Mexipedium so that Phrag and Paph could be better held apart based on morphological evidence; this was based on both DNA and morphological data. I later sunk Mex and Phrag into Paph as an alternative, should others view the characters I used to separate Mex, Paph, and Phrag inadequate. Both classifications recognize the same organismal groups - only the hierarchical levels of the names are different.
So, like my Mex example, I say let the taxonomists and nomenclaturalists not take themselves and their experience too seriously! If you like, Mex can still be a Phrag, but then my *opinion* is that you lose key characters for identification -- though you lose *nothing* in terms of a tracing of evolutionary history.
If non-objective techniques for obtaining data were not available today for mammalogy, to give an example, standard taxonomy would probably not include a group that includes hippos and whales.
Best wishes,
Vic