Probably most regulated industries. I am in medical devices and you can easily make it to 65 with dignity.

because fail-forward works so great with implanted defibrillators, I presume?

/s

“Move fast and kill people”

That's my main issue with SV-mentalities moving into the safety-critical space like autonomous driving

I have mixed feelings. On the one hand you don’t want to kill people but on the other hand a lot of the big players in regulated industries are stagnant, complacent and inefficient. A good example would be SpaceX. Compared to their competitors they move faster, are innovative, cheaper and despite that they are reliable when it counts . Most of their failures were well calculated risks that helped them move faster. I am in health care and this industry also desperately needs new players that take more agile approaches and not the super expensive super slow approach most established players are taking. They are way too complacent.

I agree with you that the old corps in these industries are stagnant, but also recognize when there's a gap in the way new entrants approach safety-critical problems.

One example is the Uber incident that killed a person. If you read the NTSB report, it seems like they intentionally programmed a delay into a safety critical system in order avoid nuisance braking. From a safety-critical standpoint, deliberately slowing the action of a time-sensitive safety mitigation is big error in engineering judgement. Imagine if I reprogrammed your home smoke detectors to wait for an hour to determine if there's really a fire because you were annoyed that they went off every time your baked. My suspicion is they did this because their software wasn't really ready but they wanted to push it to market for testing.

Another example that has had some controversy in the industry is SpaceX using only touchscreens. I don't know where they settled, but a few years back they were adamant that they don't need physical controls. We've all probably experienced issues where a touchscreen doesn't respond as intended. Automotive manufacturers have walked back this idea on a lot of their models because they recognize all the additional failure modes that touchscreens introduce.

My worry is that sometimes "innovation" is really just forgetting (or not knowing) the lessons learned from the legacy players.

The traditional players like Boeing have had their own problems, but IMO they are more due to management than engineering. With the 737MAX, the engineering processes were in place. They had a hazard analysis that said MCAS shouldn't be a single-point failure; it seems they just didn't follow their own engineering processes due to budget and schedule pressures. That's a different animal than not having the safety-critical culture to define those processes in the first place (but not any less egregious).

Sounds believable, but how many people successfully transition into those?

It depends on what you mean by "transition".

It's probably hard to transition from a software engineer to a licensed structural engineer or the like.

But it's much easier to transfer into a software or software-adjacent role related to the field. Those fields have lots of opportunity for controls engineers, or data acquisition system developers, or data scientists, or even developing home-grown applications. The flip side is a lot of bad software practices are tolerated because those running the show just don't know better or don't care because it's not in their wheelhouse.

Sounds a bit like they might put you in the IT rather than the engineering department if you come in as a SE. Control is far outside the ballpark of a typical SE, unless you happen to have studied EE (which is arguably just another one of the 'hard' engineering fields). Sounds like short of having a relevant degree you won't be an engineer in those areas, and very few people get those mid-career.

An area that people do transition into and that almost puts a premium on plain age is therapy/counselling, but that's a far shot from technical fields.

Like I said, it's anecdotal, but I've known people who've transitioned to those fields. There is sometimes an issue of conflating IT to CS, but the converse seems to happen much more in my (again, anecdotal) experience. In that case a mediocre engineer with some software experience is crammed into an IT role because the engineering team doesn't really want them. If a SE adds a lot of value to an engineering team they are usually very reluctant to give them up.

what about someone in his/her 40s with no previous experience in the field?

I wasn't necessarily saying they have to make a complete career switch but rather find a niche within those fields that can use their software experience. For example, building a pressure systems database that isn't god-awful would be useful for oil/gas or firms that do aerospace research. Knowing the mechanical side would obviously be a huge plus, but it's also possible to lean on the mechanical engineers who are the system experts. Same thing with managing environmental requirements, or building automation/energy systems, data acquisition for medical devices, or software quality testing etc.