Lets just take ourselves on a flight of fancy for a moment. Imagine you won a government contract to deliver an enterprise grade fibre backbone to a country. Backed with sizeable European funding - a major provider of infrastructure services spawns a child process, and the work is completed.
The rather dandy build is indeed a shiny thing - and lowe - a completely redundant circuit to form a loop should there be a failure on the core fibre. Hardware is carrier grade. Costs are high. Service is bespoke, quick, two gig bearers delivering us two patch leads 90 miles to our data centre routers.
Imagine how it would feel if that said enterprise service had to turn off for a bit while it rebooted after some updates.... as a single point of no-there-is-no-plan-B-we-have-to-turn-this-off failure. When you / govt. are wooing DC builders to hang entire facilities (of real scale) off of it... that... its not going to cut it chaps. Seriously.
Even with the flexibility of two feeds. Terminating in two end points. No that’s right - one chassis.