At the moment, one of the blockers for a larger Growth deployment is the low number of translations being made. Translating an extension that is not visible for experienced users is not perceived as a priority for translators.
We need to figure out ways to have Growth features being translated for sure.
Options can be:
Is this a dupe of T273635: Check if it is possible to import machine translated strings to translatewiki.net? Feels very related at least.
There is https://watch-translations.toolforge.org/, which can be used to monitor translation completeness.
It is related, but different, since it can cover other ways I'm going to explore, such as pinging projects on translatewiki, have a translathon, finding a budget for paid translators, etc.
I'm claiming the task since I actually started to work on these different options.
Note that both me and @revi rated external paid translators as performing an unsatisfying job, mistranslating a lot of Wikipedia-related terms, and sometimes even terms of everyday use. I would tend to trust paid translators from the community better (similar to the ambassadorship idea).
Echoing Martin:
External Translators do not understand Wikipedia-specific slangs, and even make a basic grammar mistakes, which make me suspect they are using Korean-as-a-Second-Language translators. Please say no to external translators. If you need to pay someone, please pay community people.
I remember your concerns from back in the days. Maybe a good glossary would help.
Think about the big picture: what would be the best? A poorly translated interface, or no translation at all? The first option seems to be the best one.
Community people would be the best option, but we need to find them, provide a contract (which can be complicated for many countries), etc. This is not an issue with translation compagnies. We need to find a way to balance the pros and cons for each solution.
At the moment, the average translation rate is 56%. Communities we engage with translate the entire interface or don't do it at all. There is no middle point.
We will think about this task later, for instance to update existing messages when we will have new tools for newcomers.
Translating the interface is an overall issue for, well, everything. If someone has a great idea that would solve the issue globally, feel free to help!