You can’t refactor reliably without tests. I started writing tests for Vassal more than a decade ago and made little headway. Almost nothing was in a testable state already, so would have required refactoring prior to having any tests, putting us at risk of breaking things without knowing it. In many cases, I couldn’t establish conclusively how things were supposed to behave, so couldn’t write tests even when things were testable. Finally, the way that custom module code was allowed access to the library code made nearly everything a public API, so it’s extraordinarily hard to refactor anything in a way that doesn’t break modules.
I’ve already been down this road. It’s a dead end.