I work on embedded systems so userbase isn't really a concern.

However we have > 10 million lines of C++, and the change would have been a major refactoring of most of them.

So I'm assuming this decision was made quite a while ago? Years, maybe even over a decade?

It took a decade to find the data to show that this decision we made 10 years ago was bad - for those entire 10 years we all thought it was good until we we looking at a performance issue and realized the roots case what a decision made in the early days had an unexpected side effect.

Congratulations on your success. Legacy code is never easy to work with, but it's a good problem to have because it means you survived where hardly anyone does.

10 years ago this was a greenfield project that was just getting the first release. Today it is a legacy project that is bringing in a ton of $$$.