Most modernization frameworks assume, implicitly, that there's a point where the old system can be turned off and the new one turned on. For the systems that move power, water, freight, and emergency communications, that assumption doesn't hold. There is rarely a maintenance window long enough, or a tolerance for risk high enough, to make a clean cutover realistic.
That constraint changes the entire shape of a modernization program. Instead of a single migration event, the work becomes a long sequence of small, individually reversible changes, each one validated against the live system before the next one is allowed to proceed. It's slower than a green-field rebuild, and it has to be, because the cost of being wrong isn't a delayed launch. It's a real outage affecting people who have no alternative to fall back on.
The operators who do this well tend to share a specific discipline: they invest as much in the rollback plan for each increment as they do in the increment itself, and they treat the existence of a tested rollback path as a precondition for proceeding, not an afterthought to document once the change has already gone in. That discipline is expensive in the short term and is the entire reason these programs don't end up as cautionary tales.
There's also an organizational dimension that's easy to underweight: the people who understand exactly how the legacy system actually behaves, including the undocumented quirks that any system accumulates over decades, are often the same people whose roles are most directly affected by the modernization. Keeping that knowledge engaged through the transition, rather than treating it as something to be replaced as quickly as possible, is usually a bigger determinant of success than the technology choice itself.