[crazygringo]

That was my first thought as well.

On the other hand, it's hard to make these kinds of judgment calls when you're talking about a one-off piece of equipment that's only going to go through this particular testing cycle a single time.

In computing, there are a lot of similar "one-off" operations -- something you to do to the prod database or router config a single time as part of an upgrade or migration.

Sometimes building a safeguard is more effort than just paying attention in the first place. And while we don't always perfectly pay attention, we also don't always perfectly build safeguards, and wind up making similar mistakes because we're trusting the faulty safeguard.

In circumstances like the one in the story, the best approach might almost be the hardware equivalent of pair programming -- the author should have had a partner solely responsible for verifying everything he did was correct. (Not just an assistant like Mary who's helping, where they're splitting responsibilities -- no, somebody whose sole job is to follow along and verify.)

[Retric]

“One off” is never just a one off it’s always part of a class of activity such as server migrations etc. Just paying attention guarantees eventual failure when repeated enough times.

This may be acceptable, but it comes down to managing risks. If failure means the company dies then taking a 1 in 10,000 risk to save 3 hours of work probably isn’t worth it. If failure means an extra 100 ours of work and 10k in lost revenue then sure take that 1 in 10,000 risk it’s a reasonable trade off.