If understanding is also the only metric then a hard g would still be preferable. Not only does a written g tend to make people lean to a hard g in my experience, but there's more words that could be mistaken for a soft g pronunciation.

What? That's just a silly claim, the word "gift" is generally pronounced [gɪft̚] with the /t/ having no release, often the last consonant isn't even perceived by speakers, if anything that is extremely easy to mix up with "gif" using a /g/ as opposed using a /dʒ/, compared to any other words (well I guess there's "jif" the peanut butter brand?). You make a bad argument.

Also yes, if someone pronounced or used a word one way and then went to some theoretical place where everyone else pronounced or used it in a way where it becomes mutually unintelligible, then yes you WOULD be saying it "wrong" if you insisted on pronouncing it in a way nobody can decipher it, if you can call anything in language "wrong". French speakers can't just go say shit to Sicilian speakers and expect to be understood.

But no, there are no rules about word construction or pronunciation. The closest thing we have to "rules" is loose standards that people commonly us. And in the context of this conversation, most English standards don't invoke any sort of phonemic spelling like e.g. Spanish or French or Polish or Korean or whatever. There are no "spelling rules" that dictate that a certain sequence of letters or words has to be pronounced a certain way regardless of context, even according to standards of English. None of that "exceptions" bs, Modern English spelling is mostly based off of a writing system of a language that Modern English speakers wouldn't even understand, and as such there are only a few sometimes-consistencies-ish, like using certain constructs to differentiate lax vs tense vowels like doubling the following consonant letter vs appending an "e" at the end, when applicable. It's just infeasible due to the history of the writing system to apply a consistent convention for phonemic spelling without reforming the entire orthography.

This is opposed to, say, French, in which standard spellings have actually consistent throughout the entire language rules for how a certain combination of letters is formally pronounced (regardless of how much French speakers like to claim their spelling is nonsense), sometimes with secondary/uncommon pronunciations, and with exceptions to those rules. And consistent rules for phenomena like liaison. And applying those rules, you can systematically pronounce a majority of words accurately even if you've never encountered the language in your life. Here's a table just for fun: https://en.wikipedia.org/wiki/French_orthography#Spelling_to_sound_correspondences

This is not something you can do in English.

And even using the argument of standards, the most common descriptions of Standard English (e.g. Oxford's dictionaries, Merriam-Webster, AHD) all list both /gɪf/ and /dʒɪf/.

Also you claim that the latter is falling out of favor, but that seems to have come from thin air. All the resources on the matter in the first place are online polls with a small sample size and a lot of bias in terms of the location of the respondants from like a decade ago, idk how you determine that one is more popular than the other in a way other than "I hear X pronunciation more than Y". The fact that this argument is seen all over the internet and is extremely contentious should be proof enough to show you that that claim is fallacious.