The threads below contain a mishmash of information (and some misinformation and opinion, nothing new there), so maybe a simple overview would help. Not many people are interested in the details of taxonomy and classification, and fewer still actually publish in that field, but we all use the names. Where do scientific names come from, and why do they sometimes change?
Scientific names are formed in Latin (or Latinized versions of Greek, or any other language) for two reasons. First, when the need for a standard way to refer to species became necessary in the mid 1700s, Latin was the only shared language among European countries – no matter if your native language was Italian, German, Swedish, English, etc., anything that needed to be widely understood was written in Latin. In other words, Latin then had the status that English has today. Second, everyone accepted that the only way to avoid confusion was to have a set of internationally-accepted rules about how organisms were to be named and referred to. This was standardized around a hierarchical system based on a listing of all species then known, authored by Karl von Linne (or Linnaeus, in Latin) in 1755-57. Those principles grew into what is called the International Code of Zoological Nomenclature (there are also codes for plants and single-celled organisms).
The basic principles of the Code now form the oldest international 'treaty' in modern history – it has been supported by almost everyone for 250 years, despite language barriers, wars, and all manner of political and social disagreements among nations. There are always proposals to change it or dump it, but thus far none has succeeded, probably for the simplest reason – it still works.
The original rationale for the Code was nonevolutionary (100 years before Darwin, after all), grouping organisms solely by their by structural similarities, ranging from "kinds" (= species) into increasingly inclusive categories (genera, families, orders, classes, etc.). This nested-boxes approach works just as well in an evolutionary framework, where similarities in structure are regarded as evidence of relationships via shared ancestors, and a useful classification is now considered to be one that accurately reflects the evolutionary history of the group.
"Priority" is a key feature in scientific names – the earliest name that meets the conditions listed in the Code is the one we agree to use. This does perpetuate some errors. For example, it is accurate to have a species called V. komodoensis, but no longer as informative to refer to something as V. indicus, when it does not occur in India. This is an historical "error", because all of the SW Pacific islands were termed the East Indies at the time. There are also just plain mistakes, where the person naming a species had incorrect information about where his specimens came from. That's a cost of priority, where the earliest information is not the best, but it is the standard that everyone accepts. Animals may be named after their distinctive features (V. glebopalma, V. prasinus), after the places they were thought to occur (V. bengalensis, V. cerambonensis), or after people who had a lot (V. mertensi) or nothing (V. teriae) to do with the biology of monitors.
Established scientific names can change for two general reasons, legalistic and biological. It sometimes happens that an older name has been overlooked, and by the law of priority should replace the name then in use. Where such an 'established' name has a large literature, and it can be argued that changing it would be disruptive, there is an international commission that can conserve the name in general use. This happens quite often, for example when it was discovered that the original specimen of V. gouldii was in fact a V. panoptes – in the interests of stability the commission ruled that the names gouldii and panoptes should remain associated with sand and argus monitors, respectively, even though Storr erred in 1980 in thinking that argus monitors, not sand monitors, were the unnamed species.
Biological reasons for name changes have to do with cumulative information from new research. For example, Mertens described glauerti, scalaris and similis as three subspecies (geographic races) of V. timorensis. Storr later showed that glauerti was distinct, as were the Australian and Indonesian 'Timors', now known as V. scalaris and V. timorensis. With further work (based on better series of specimens, and more detailed databases), Laurie Smith is finding that scalaris contains at least five genetically distinct species (which will be V. scalaris, V. similis, and at least three additional species that will have brand new names, because there are no prior, validly published names for them). This is "splitting", whereby one named species is found to be two or more. To achieve general acceptance it is usually necessary to show that these species occur together without interbreeding (as is the case, for example, with V. gouldii and V. panoptes), or are so genetically distinct that interbreeding in nature is very unlikely. Disputes arise when these criteria are relaxed, as for instance in applying various new names to island populations of timorensis or prasinus – some people are convinced, others are not, and they fight about it.
Alternatively (and less frequently, in this era of gene sequencing) evidence arises that two named species are geographic or color pattern variants of the same genetically-defined species (for example, Bell's lace monitors). In that case, which involves "lumping", the older name is adopted for the entire unit.
Changes in the allocation of scientific names are not done for trivial reasons – instead, the goal is to have a classification that reflects the genealogical relationships, or evolutionary history, of the group. This is necessarily an ongoing process – biologists find animals that are clearly new species (such as V. mabitang), or new specimens and types of information give a clearer picture of relationships (such as understanding that what we once called V. indicus on New Guinea is actually made up of four quite different species [doreanus, finschii, indicus and jobiensis]). It is progress, mostly. This is not to say that all proposed changes are well-founded – there is some sloppy work in the monitor world just as anywhere else, and we typically look to "acceptance" by the larger community as the correcting factor. Strange new names appear fairly often (for example, V. irriwadicus, which is actually V. bengalensis nebulosus), and may lead to some confusion for a time. If they are not accepted, these names are relegated to being listed as invalid names beneath the accepted species name in a section termed a "synonymy" that appears in more formal species checklists and accounts.
Along the way, I think that this answers most of the questions that Frank raised below. Scientific names are as stable as they can be, given that knowledge about the animals themselves increases. The fact that names sometimes change simply means that we are getting closer to an accurate understanding of diversity and relationships. Names often do not reflect the animal because the first person to propose the name "wins", whether the name is appropriate or not. We need scientific names for the same reasons we always have – so people who speak Japanese, Hindi, Swahili, Hebrew and Farsi know what I mean when I refer to Varanus brevicauda, and so do you. It's also true that the animals don't care – what the heck, Tyrannosaurus rex never knew what it was called, right? But if you want to refer to it, we seem to need something to call it.
Last comment, on "cladistics". This is a philosophy and set of practical rules about how one does taxonomic work, which focuses on using shared, derived character states to reconstruct evolutionary histories. The basic idea is that if three species share some feature (of morphology, of a gene sequence, whatever) that is not found in other monitors, then we regard those species as having had the same ancestor. If two of those three share another feature that is absent in the third, we term that pair sister species, which are said to share a more recent common ancestry than either does with the third. As long as the characters we use are informative (they vary among species, don't vary within species), you can build a cladogram, or tree, that is the best estimate of genealogical relationships within the group.
This method and philosophy has largely replaced other concepts of analyzing evolutionary relationships, starting in the early 1970s. As a method it is logical and practical. In my own opinion, the cladistic philosophy goes a little haywire when it attempts to modify the Linnean system of describing relationships in a formal classification. Some cladists feel that every branch point in a tree (cladogram) requires a unique name: in Varanus, a single genus with about 60 species (and 8 subgenera), this would involve coining on the order of 25 additional scientific names that refer to different nested groupings of species. Apart from being unwieldy, the memberships of many of these limbs, branches and twigs are still subject to change as additional species are added, or as different characters or genes are included in the database. Now that would get confusing fast, because you would need to know which author's concept of 'Kakamamievaranus' somebody was using.
