In Switzerland, arguably the world's experts in tunnel-boring, one death per tunnel kilometer is considered acceptable (and it is often less than that).
The much easier Channel tunnel has had around one death per five kilometer.
The Japanese Akashi Kaikyo Bridge prominently has had no construction deaths, which was considered extraordinary at the time. The Humber Bridge, which was the longest suspension bridge before Akashi Kaikyo, did cost four lives.
Construction is a dangerous business. You operate heavy machinery. You move large amounts of heavy material. You deal with power tools. Death is always a possibility - and that's something we IT folks tend to forget in our clima-controlled environments with our cozy desks and expensive chairs.
The US Death Rate per Capita is about .8%, so let's start there.
How many is it acceptable to die on the job? Well, there's hard stats on that - About 1M people work in construction[1], and about 1000 of them died on the job last year[2]. That'd be about .1% death rate on the job. Is that good enough? Probably not, but it's where we're at and what society accepts.
If 2,000 people worked on the Golden Gate Bridge for five years, it'd be an average construction project. It was a massive undertaking, but apparently poorly documented - There's no record of worker counts[1], but we do know the number of deaths - 11, of which 10 were from a single failure of a gantry.
Obviously more than zero, else we wouldn't attempt to build any bridge (or do anything else, like drive cars or fly planes) lest someone dies in the process...
What's acceptable (i.e. It's ok if 1 person dies on this project) is different than what's the potential risk (i.e. There is a risk that 1 person might die on this project).
If you're working on a project and 5 people have died so far and you've still got a lot more to do I think it's pretty reasonable to expect that more people will die. So, if you're going to continue the project you're considering it acceptable that somebody else will die as well.
Sure, you shut it down for awhile and then just let it start back up again and after the next death shut it back down and rinse-repeat. If it was the 1930s you wouldn't have to bother taking the break in between deaths but the rinse-repeat means we have a non-zero tolerance for death.
if 5 people have died the same way so far, modern workers would not find it acceptable that more of their friends will die from the same unaddressed cause before the building is done, and thus would not, in fact, continue the project until something is done about it.
And further there are trade offs. I don't know the trade offs of bridge building is but the example i've heard in the past is, it's possible we could make it so no one ever dies in a plane crash. To do that with today's tech might make flights cost 2x, 3x, 4x more. Flights cost more = more people choose to drive (say SF to LA or LA to Vegas or DC to NYC). More people drive = more deaths.
That isn't to say you shouldn't make things safer but you have to take into account is your change a net positive over all. I have no idea what the trade off is for bridges. The bridge arguably increase commerce which means its completion created jobs. Poverty is a top killer of people so more jobs = less death.
That said, one thing I'm happy to see on construction sites, at least for buildings, they put up fences on every floor so it's harder to fall out. My guess is that really only became possible when they could create the fences out of plastic.
I think plane crashes are not a good example of your point. Every time a crash happens, a massive post Mortem tends to occur to determine what needs to be done such that that specific failure mode will never happen again. Pilots have checklists on checklists on checklists that they must tick off during each procedure of the flight, with each point written in the blood of previous crashes. There isn’t discussion generally as to whether or not adding new checklists, the two pilot system, or other safety improvements as a result of this would make flights more expensive: hell, an entire new model of plane was taken completely out of the skies during a recent investigation as to a fucked up sensor.
Plane crashes might have improved but plane flights have not. I'm old enough to remember flying in the 90s. I'd be willing to play Russian roulette if I could show up 15 minutes before an international flight today and still make it on the plane.
You could also make commuter driving (near) zero-death now. It would be a huge one-off payment for a national - possibly international - navigation and power grid. But once built it would be a huge cost saver, because automated traffic flows could be optimised across the entire system.
Although in fact remote work would be even better, for those jobs that support it.
Meanwhile jobs have become poverty traps, so they're not enough on their own to avoid poverty and death.
Ultimately it's not about building things or the cost of highly visible prestige projects. It's about systemic rather than fragmented thinking. Capitalism tends to the latter, because an airy wave of the invisible hand is supposed to somehow optimise everything.
This is magical fantasy. If you want intelligent culture-wide systems you have to get your best people to design and build them. And they have to consider long-term systemic outcomes as much as short-term goals and projects.
This culture is very, very bad at that. But it's also unrealistically convinced it's very good at it.
