Ports of massive legacy codebases are possible and they happen. They can be extremely difficult, they take will and effort but they can get done. The idea that you have to slow down the development of the language standard for people who won't port to the new version is weird- Those people won't be updating compilers anyway.

How do I know this? I migrated a codebase of about 20m lines of C++ at a major investment bank from pre-ansi compilers to ansi conformance across 3 platforms (Linux, Solaris and Windows). Not all the code ran on all 3 platforms (I'm looking at you, Solaris) but the vast majority did. Some of it was 20 years old before I touched it - we're talking pre-STL not even just pre ansi. The team was me + one other dude for Linux and Solaris and me + one other different dude for windows, and to give you an idea the target for gcc went from gcc 2.7[1] to gcc 4[2], so a pretty massive change. The build tooling was all CMake + a bunch of special custom shell we had developed to set env vars etc and a CI/CD pipeline that was all custom (and years ahead of its time). Version control was CVS. So, single central code repo and if there was a version conflict an expert (of which I was one but it gives me cold sweats) had to go in, edit the RCS files by hand and if they screwed up all version control for everyone was totally hosed until someone restored from backup and redid the fix successfully.

While we were doing the port to make things harder there was a community of 667 developers[3] actively developing features on this codebase and it had to get pushed out come hell or high water every 2 weeks. Also, this being the securities division of a major investment bank, if anything screwed up real money would be lost.

It was a lot of work, but it got done. I did all my work using vim and quickfix lists (not any fancy pants tooling) including on windows but my windows colleague used visual C++ for his work.[4]

[1] Released in 1995

[2] Released in 2005

[3] yes. The CTO once memorably described it to me as "The number of the beast plus Kirat". Referring to one particularly prolific developer who is somewhat of a legend on Wall Street.

[4] This was in the era of "debugging the error novel" so you're talking 70 pages of ascii sometimes for a single error message with a template backtrace, and of course when you're porting you're getting tens of thousands of these errors. I actually wrote FAQs (for myself as much as anything) about when you were supposed to change "class" to "typename", when you needed "typedef typename" and when you just needed "typedef" etc. So glad I don't do that any more.