It is often quite good, but it could be vastly better. I remember helping to translate an article into English. It was written mostly by my friend, whom I know to be impatient with phrasing, and in the course of translation he often changed the original text, added caveats etc. The gist of his meaning remained, but it was framed very diffently. Often, however, reviewers are too hurried and point out only the worst mistakes or unclear places. Perhaps it would be useful to have a 'malicious revier', a person unversed in the given field but fluent in data presentation, send his comments to the author before the article is shown to domain experts?..