But how is syntax based version control ever supposed to work when we have so many languages that literally parse differently depending on dependencies that aren't even in the codebase? Let alone ones that easily take hours to compile even when they are in the codebase?
Sounds impossible in practice unless we restrict ourselves to sufficiently context free languages.
You create a language geared toward syntax-based version control which doesn't do that, and then relegate syntax-based version control the ghetto based around that language (not offering it as a general-purpose tool for programming in anything).
Languages with user-defined subgrammars, like Raku (Perl 6).
Languages with declaration-dependent ambiguity resolution, like C++ (https://en.wikipedia.org/wiki/Most_vexing_parse).