I don't buy the part about asymmetry of switching costs. It's possible to argue that asymmetry exists in the opposite direction: each employee has one job, so switching jobs is time-consuming and risky; Google has thousands of engineers who have had the same training/experience on internal tools/standards, so it's easy to replace a single engineer when they leave a project.
The other point I made is analogous to Baumol's cost disease except that, instead of applying to different types of job, it applies to different types of employer.
- developing a training curriculum or materials
- designing operational processes
- developing a brand identity
The difference with software is that the software kind of runs itself, without much need for humans. The examples I gave above are also 'create once and execute many times' things, but they require humans to do the execution.
They fired me because they knew they couldn’t afford to keep me and if I was courting these companies, they knew I would get an offer. Which was true.
And if they knew that you were going to get an offer they couldn't match, then obviously you were worth more than what they were paying you. In other words, a great deal for them. Why would they cut that any shorter than they had to?
Was it just good old fashioned spite, cutting off their own nose?
If tech salaries were risk premia, you would expect the expected value (weighted average over all foreseeable futures) of compensation packages at startups to be higher than those at established tech giants. This might be true, but it doesn't seem obvious: there are often comments on HN doubting the claims of high packages at tech giants, suggesting the commenters aren't getting paid more at their more risky (p(fired | laid off | shutdown)) jobs.
Software engineering salaries are high because there's competition for the 'best' software engineers. In a competitive hiring market, the absolute upper cap on compensation is the marginal contribution of an additional engineer, less some percentage to account for uncertainties. That marginal contribution is high at the largest tech companies, so salaries are high. Other companies can either pay high, too or, if they can't use a software engineer as efficiently (e.g. because they don't get as much leverage from software, or if their engineering is poorly managed), they won't be willing to pay the market rate.