> for real companies
I wouldn't consider those entirely equivalent sets. I imagine plenty of startups probably don't fall under the criteria you would consider "real companies", or at least not in the beginning before people have a chance to mature into their roles or flunk out of them.
> the economics of ruthlessly withdrawing bounties don't make sense
The economics of something and how people try to justify it or let their own egos get in the way often don't match. I mean, I still have to kick myself sometimes because while I work at a small company, agonizing over a couple hundred dollars a month in service fee differences is not a good way to spend my time given my hourly rate and the time a more expensive option might save if it does what it says. Ingrained thinking can be hard to overcome.
These days I generally just sit on issues. The work involved in putting together a bulletproof report that can be understood by whoever reads the security alias (could be a security engineer, could be a PHB, could be /dev/null...) is just too high to do for free.
There are some patterns where I've seen people not get paid just on general principle; for instance, people find systemic issues and, rather than disclosing the root cause, try to claim bounties for every instance of the flaw (you'll get paid, but not for every instance). It's possible that naive development teams sometimes get this confused, and, for example, consider "all XSS" to be a single systemic bug.