I’m undecided as to whether researchers should publish in their native language or English. If my native language were not English I probably would publish most of my papers in my native language in a local journal and publish short English papers on my work in international conferences. I’d do that because scientific translation is too rare at the moment for non-English language research to be visible internationally, even for abstracts.
Scientific translation is an interest of mine. In my own research I try hard to do a comprehensive literature review and this includes searching the non-English language literature. I have a lot of experience identifying whether a paper has an English translation and tracking down English translations. See this long Stack Exchange answer I wrote. Scientific translation seems to have died down considerably since the late 1980s due to the end of the Cold War. This is a shame as good research not in English still exists.
I’ve published translations of 7 articles which I translated myself using Google Translate (with heavy editing). I published the translations in my university’s open access repository. Since I don’t know German or Russian, the original languages of the papers, I marked sentences that I am unsure of the translation of. People like to complain about the quality of automatic translation software, but for the most part the results are intelligible. Translation this way is tedious, and in my experience very few academics think it is worthwhile, but I believe it’s worthwhile for me. I’ve avoided duplicating quite a lot of work through comprehensive reviews and translation as needed. (The fact that much of what I do is typically viewed as unproductive to traditional academics has helped motivate me to go independent.)
English abstracts of non-English language papers are common in some fields, e.g., math has Mathematical Reviews, which is excellent and fairly comprehensive. Russians have Referativny Zhurnal for nearly all fields, which was quite good until the end of the USSR.
In my own work I have found English abstracts to be useful to identify non-English language works to translate. While this is not perfect, it is much better than nothing. And, in my experience the main problem is the abstract not describing something of interest, not the accuracy of the translation.
With that being said, translation is a hard problem that requires both domain and language knowledge. It is very rare for someone to have both! Typically professional translators have only the latter, and this frequently leads to literal translations which may not make sense in English. Often the reader can figure out what was intended, e.g., a recent translated title I encountered is “The Problem of Closing in the Theory of Turbulence”. Readers familiar with turbulence would know immediately that the translator should have used the word “closure” instead of “closing”. I could probably make a long list of similar issues, some of which are more consequential. But the typical case is like the example I gave, where the error is obvious to someone with domain knowledge.
I’ll agree with @florence.cotel in that it would be best to have a relatively small number of people focused more on communicating the results of research to an international audience than for everyone to learn English.
As a side note, one of the more famous independent researchers, Julian Barbour, worked as a scientific translator.