This is a very difficult question to answer, since it deals with future developments in the language. Unlike other languages (for instance, French), there's no regulatory authority for the language.
Even in the case of l'Academie Francaise, they are largely ineffective in determining how language is used in informal contexts, though I'm unsure of their effect in institutional settings - they may attempt to do so through scholastic, academic and governmental instruments, which might actually be effective.
In English, plurals are formed in a few ways (and there's confusion about words, even those that have been in the language for a long time). The main issue is that there are really no defined procedures for determining which way a loanword will pluralise. Some of them pluralise according to the rules in the donor language, others pluralise according to productive, regular English rules.
As for your examples, I would likely pluralise them as:
- /pelmen/ ~ /pelmenz/
- /strelets/ ~ /streletsez/
However, this would only be the case if I had never seen or heard them before and had no native speakers around to ask. Why? Because while they're still foreign words, used in English, meaning I'd want to use the foreign pluralisations. This depends on the individual speaker, though.
Once incorporated into the language, it's difficult to guess whether native or foreign rules for pluralisation will apply.
- orangutan, borrowed from Bahasa Melayu, would be pluralised as orang-orang hutan, as reduplication is a productive process in that language. In English, however, the pluralisation is simply orangutans, honoring English processes.
- beau, from French, however, pluralises both ways - as beaux and as beaus, as in French and English respectively.
The short answer:
- I would ask native speakers what the Russian plural is, but mainly for my own knowledge;
- I would pluralise them as English for audiences unfamiliar with the native plurals, and;
- I would pluralise them as Russian for audiences familiar with the native plurals.