The article mentions PSR -- precision scheduled railroading -- or some such as at the core of all this.
Okay, it's a scheduling problem. It's an optimization problem. It's possibly to a significant extent an optimization under uncertainty problem. Oh, and it's an optimization over time problem.
Did my Ph.D. dissertation in one of those, not for trains but for airplanes. Oh, for math majors in the audience, yes, did use Fubini's theorem (the core of the proof that the work really was in a very general sense actually optimal) and measurable selection.
Such problems can be challenging in practice for business operations, for the math, and for the computing.
And, right, the problem is no doubt, oh, I mean, even quite simplified versions of the problem, in NP-complete or if don't like that fact then can be challenging to get even approximately optimal solutions.
But, however look at the complexity of the problem, I'd
Guess: Including a 7 day paid sick leave should be possible to include in the problem.
That is, just because are attacking a challenging optimization problem, there is likely no fundamental or even very important practical reason can't also include as a constraint, feature, property, requirement, whatever, something like 7 days of paid sick leave.
Besides, we're talking about scheduling workers to run freight trains, maybe 100 cars long, with several big Diesel-electric engines, across major parts of the US. Uh, just the Diesel oil costs what? And in comparison for that trip the workers cost what? I'd guess a small fraction of even just the Diesel oil. Then there is the capital cost and maintenance cost of all that equipment. E.g., the track has to be in really good condition or ... don't want even to imagine what the costs could be, i.e., the term train wreck is used broadly as a really big disaster. And just a little maintenance on one of those Diesel-electric engines costs, how many hours of worker pay?
My nose smells more politics than business, applied math, computing, technology, etc.