In my personal experience, the older the legacy system, and the more embedded it is in your business ... the harder it is to replace.
But if it's that old, it's probably also massively underdocumented (if at all) and so if something unexpected happens, your ass is still hanging out the window. Producing the documentation of what was actually done is at least as valuable a part of a replacement project as the change to the new system, as it should allow someone to start looking at which parts are required, which parts are emulating interfaces (from both sides, usually) that could be de-layered for improved performance and capabilities with no down-side, and which parts are just dumb holdovers from a few systems ago that nobody needs any more at all.
Just because something is painful doesn't mean you can get away without doing it.