The all-at-once replacement is still the most expensive way to modernize a core system, and the most likely to be cancelled at 70 percent complete with nothing in production to show for the spend.
The alternative is unglamorous and reliable: put a stable interface in front of the legacy system, move one capability at a time behind it, and route traffic gradually. Each slice is independently valuable and independently reversible. The legacy platform shrinks until switching it off is an operational task rather than an event with a war room.