In many cases, things like customer lock-in, natural monopolies or intellectual property catalogs serve as significant barriers to entry even for well capitalized new businesses, and these barriers can more than outweigh the costs of having half your employees being essentially useless. Finally, an argument could be made for resilience in staffing. The very leanest organisation would have bus factors of 1 for every single job and the first sick employee means problems. Even raising the redundancy to "only" two people per critical function means that you just created a ton of bullshit jobs. (since most of these people will be severely underutilised)
The market is a dynamic system, constantly trying to reach some kind of equilibrium (defining which is left as an exercise to the reader). As such, it evolves through time, and has some degree of inertia. It's perfectly possible - and in fact, common - for such dynamical systems to oscillate and orbit the desired state without ever reaching it.
If you took a market as it is today, and have everyone just do whatever it is they're doing, then it's likely it would've reached some stable state. But in reality, market participants constantly change what they're doing, responding to and trying to predict what everyone else is doing.
(Especially that last part is one that's damning to any kind of predicting.)
In such conditions, bullshit jobs can persist, because they appear faster than the market is optimizing them away.
constantly trying to reach some kind of equilibrium
apologies to ask such a naive question, but im trying to understand, what is the basis that it is thought 'a market is "trying" to reach an equilibrium'...maybe im too hung up on the semantics/wording, but im genuinely curious because in my mind "the market" isnt really "trying" to reach anything.... maybe its agents are (gain a profit, hoodswink someone, make a fair exchange, etc etc)... so in aggregate there may be some result over time, but i find it hard to imagine its trying to reach an equilibrium per se...?