I can understand not liking how bureaucratically onerous it is, but we don't really have a choice. It's imperfect but it's the only way we've found to manage highly complex large scale engineering projects. You certainly aren't going to build something like JWT with a just YOLO agile process unless you're literally willing to fail like 50 times before you get one right (aka the USSR approach to rockets).
The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity now it allows a bit more virtual trial and error in the process.
>The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity
I agree to an extent, but there's also a risk of simulation breeding a false sense of security even when simulations are conducted well. The investigation of the CST-100 "anomalous" test flight had 21 findings related to software simulations and testing, some related to lack of fidelity. Not that fidelity wasn't possible, but it has some overtones of the Ariane 5 software issue in that there was a lack of integration testing within the different software components.
I agree, but the problem with this case was that simulations led to complacency about not thinking they needed integration tests. That's much rarely in the hardware domain. It's not really meant to be a critique of simulations but rather how we use them.
The upside is automation is making it go smoother and be less burdensome, because simulated testing truly is high enough fidelity now it allows a bit more virtual trial and error in the process.