On the question of interface localization, there are several questions to be asked.
First, of course, whether we should have it at all at some point (to which I think the answer is a clear yes).
Second, if it needs to be in MVP. The starting post claimed that it is painful to add it later. However I wonder if that pain really came from not adding it up front, or rather from not considering up front that it eventually will have to be added. That is, do we really need to make the effort to build localization in from the start, or would it suffice that we make sure that the right hooks for localization are there, and that the interface reacts appropriately to strings of varying lengths. And related to it, how much different is the work between supporting localization, and supporting the later addition of localization.
The third question, which will likely only be relevant once we start to actually localize, is which localization framework we will use (assuming we don’t write everything by hand).
The fourth question (which certainly will be relevant only when actually implementing it) is on which level the decision should be made: At the instance level? At the community level (possibly with an instance default)? At the user level (possibly with a community or instance default)?