Two weeks ago on the Taxacom listserv, there was a lively discussion starting with the question, "Does the species name have to change when it moves genus?"
From a proximate standpoint, it depends on the gender of both genera, whether the species epithet (the second part of the species name) is a Latin based adjective, whether the species epithet is two or more nouns in opposition (otherwise known as a compound word), and other rules that are particular to the International Code of Zoological Nomeclature. There are, of course, different yet similar rules for plants. This is all a (sometimes not so simple) matter of reading and understanding The Code.
However, the question was posed from an ultimate standpoint. They were not discussing Chapter 7 of The Code, but rather why it should be necessary to change the names at all. Why change the gender of names when moving to a new genus? As Roderic Page pointed out, "We don't change the name of a species called "africanus" if we discover that the specimen locality was actually from Australia, nor do we change the name "maximus" if we subsequently discover a bigger species." Why not just keep the species epithet the same regardless of what ever genus we're moving it to? Wouldn't this be less confusing?
The discussion members took several different positions in their answers, and you can read those over at the link above. I agree with the way things are for several reasons. First, not changing the genus part of the name when we switch between genera is simply confusing and nonsensical. The ultimate purpose of taxonomy is a general reference system for all of biology, one that is descriptive, predictive and explanatory. The power of the binomial species name is not only the large number of possible combinations, but also that it contains information about the hierarchy in itself. We not only know the name of the species, but we also know the higher taxon that it and other closely related species are hypothesized to belong, all coming from a nearest common ancestor.
Second, we change the species epithet only in cases where it is a Latin adjective that doesn't agree with the gender of the genus. This is not true for names that are not adjectives; we do not, for example, change the species epithet of Keroplatus fasciola when it was placed in the neuter gendered genus Rocetelion, because 'fasciola' is a noun that means "little bandage", not an adjective. The reason, in my opinion, that we make Latin adjectives agree with the gender of the genus is because to do otherwise would be sloppy. If we're going to use, explicitly, this well defined dead language as part of our biological nomeclature, not just because we think they look pretty but because their meanings are relevant to the nature of the species in question, then we need to do it right.
By all means, if you don't want to deal with what you deem to be a silly anachronism, use a noun in opposition, use a word from another language, use a patronym, use a made up GIBBERISH word for all I care. But if you are going to go with tradition and use a Latin adjective, please, for the love of Carl Linneus, don't be sloppy about it. You aren't required to use Latin or Greek or even any word in any known language when you coin an n. sp. For example, generic anagrams abound, and at some point I'm going to describe just how crazy that can get.
As far as species name changes in general, we would have this problem regardless of the above issues. Synonyms and homonyms are constant taxonomic issues. Homonyms (names that are the same but actually refer to different species concepts) have to be changed to avoid confusion. Synonyms (two or more names that are different but are found to refer to the same species concept) have to be combined to reflect our current understanding of species and their relationships. These constantly wreck havoc on the stability of biological nomenclature, but they are issues that are not going away. This is a much bigger problem than genus changes and gender agreement, and if we can address this bigger problem we might as well not skimp on the little things.
Hat tip to Morgan Jackson for the link.