The only other point I would like to add is the actual internal variation that would complicate a scientific article. The dialect that is typical of my home area and the dialect that is taught at school have incompatible spelling due to pronunciation, but is verbally compatible with some amount of deliberation.
An example would be "Ke tla go bona." (I'll see you [later]; correct spelling) That's actually pronounced "Ge dao bona" in the variation. There are not enough people interested to find a way to get around the standardisation issue and, in fact, whichever dialect becomes standard is inherently a form of discrimination. This is what has happened to the Sepedi dialect of Northern Sotho. It has become the politically correct term. I've heard many people introduce themselves as a Mopedi (a Pedi person) while I would later point out to them that they are from a different dialect, which they would immediately concede: they meant it as umbrella term. In their defense, there is a difference between linguists and between actual users of a language. I guess you could say Sepedi is colonising the other dialects, though without any violence as the term might insinuate.
One reason might be that "Sesotho sa Leboa", which is the correct term for the language, Northern Sotho, is almost never used in verbal conversation. The term was almost certainly invented by linguists. The most common term would be "Sotho", but due to the confusion with Southern Sotho (a separate language) the default then instead becomes "Sepedi". It is reminiscent of "Castellano" vs. "Español".
For the curious: sheng-wen-sen-te-he-ge-lin-na-ding-si. 圣 sheng is a fully conventionalized loan-translation of the Christian word "saint". 文森特 wen-sen-te is a sound transcription of Vincent. 和 he is an ordinary Chinese word meaning "and". 格林纳丁斯 ge-lin-na-ding-si is another sound transcription, for "Grenadines".
The way to make a language fit for writing a scientific paper in it is to write a scientific paper in it, and whenever you're at a loss for words, either make them up as you go along or take them from another language. The latter option is usually chosen by people already fluent in that language, which is how English scientific vocabulary got its multiple layers of Arabic, Greek and Latin loan words.
However, loanwords can render the underlying logic incomprehensible. For example, I recently watched the Japanese series アンナチュラル annachuraru (Unnatural [Death]) on a Chinese streaming platform (because that one didn't have region locking blocking European IP addresses). When the word エチレングリコール echirengurikōru (ethylene glycol/ethane diol) came up, obviously someone had to ask to repeat it, because it's such a long and complicated foreign word. The Chinese commenters reading the subtitles were very amused by that, because the Chinese translation is 乙二醇 yĭ èr chún and has a systematic composition: 乙 yĭ is used like the Roman numeral Ⅱ or the Latin letter B in enumerations and indicates here an organic molecule with two carbon atoms. 二 èr is the number two. 醇 chún indicates alcohol. Essentially a mirror image of "ethane diol" that even someone without much knowledge about chemistry can understand.