Well, immersion, particularly in the moment when the camera is tracking in towards Michael's face (and the tension visible upon it) just before he does the hit; it loses so much of its intensity when you're trying to split your attention between the subtitles and the images above it, you know?
Isn't this a problem with any scene in any film that is subtitled? Do we not always have to divide our attention between text and performance with subs?
There are no perfect solutions. We might
(a) live in a cinematic world where everyone is always speaking our native tongue (just have characters speak English all the time),Option (a) is convenient, but ain't much on verisimilitude (e.g., Star Trek, where they journey to strange new worlds to speak English with everyone!). Option (b) is the most pragmatic/easy for foreign language films, but also seems to result in a lot of mischief, if not vandalism of the original film. Option (c) is the high-brow option, but is the most cognitively challenging. Option (d) cannot work for too long before people will just shrug and exit the theater.
(b) insert an English overdub into foreign language films, so that we can hear the words and focus on the performance (at the cost of losing the original vocal performance and having a mismatch of mouth and sound),
(c) split our attention between performance and text, or
(d) just sit dumbfounded, not knowing what is being said
That being said though, you could argue that it would've made more sense if all the dialogue in the first half of that scene (before Michael makes his decision to go to the bathroom and do the hit) had just been in English the whole time, since they switched back to it anyway at a certain point, and only then would Sollozzo switch to Italian when Michael returns in order to make a more personal appeal to him, since Sollozzo senses that they've reached an impasse. And, leaving that particular moment unsubtitled would have a dual effect, since it would have us tuning out Sollozzo's words at the same time that Michael's also tuning them out, even though he can understand everything (as opposed to before, since he was obviously listening and responding to every word, even though it was still unsubtitled to us).
Sounds plausible to me.
But again, our example is that of a short period of untranslated speech. If the Godfather were shot with 90% of the dialogue in Italian, we would have to have subtitles or a dub.
And the point in that scene, as I understand your analysis (which is rather astute), is not that this scene forces non-Italian "speakers to listen closer to keep up, or even miss a few jokes," both of which being "part of the point." Rather, we are pushed out so as to enter the emotional subjectivity of Pacino. We are being excluded, but not because it would "disrespectful" to speakers of Italian to feel as if they were not part of the intended audience. This wasn't done for a moral scruple, but rather an aesthetic result, to which I say, "sure, cool, great."