1: Many of those "overpaid consultants" were doing the work at a time when the number of people available was low compared to the number of places that needed it done, and often it wasn't just "insert a floppy" but to first work out what systems needed updating - depending on the software and hardware in use that could be a dozen very different programs (or more) and either a bios update or OS patch to work around a hardware issue.
So the company using them might have been charged a lot by an outsourcing company for what seemed like a very quick fix, but that would have been after the actual work needing to be done had been identified, the patches sourced (in some cases written specifically) and tested - the "insert floppy" stage was usually the final step in something that in a big business might have involved hundreds of man hours of planning and testing, then however long it took to update and confirm for each of the machines (in some cases hundreds/thousands).
2: How many legacy coders do you think there were at the time?
And how many of them knew the systems they were watching?
Most of those claims were in regards to systems like banking where the total number of people intimately familiar with the (often ancient) code running in the background could be counted in the low single digits for a major bank, so not only were you paying for skills and knowledge that in many cases was exceptionally rare, but at a time when many people even in low end jobs would be looking at double or triple their normal overtime rate, and a lot of those people had long since retired, meanwhile the systems they were being paid to babysit could cost many times what they were being paid in total for every minute of downtime.
It really doesn't take long for a banks system to be down for the cost in lost business and regulatory fines to hit 7 or 8 digits
From what I remember every instance of people being paid "a years salary" was for people being expected to be ready to deal with either life critical systems, or financial systems where downtime costs a fortune, and where the number of people who could troubleshoot a problem and fix it on a live system in a hurry was very limited. I remember reading stories from people involved where there might have only been two or three people alive who had worked on the original code well enough to fix it, and in a funny turn of events they could pretty much name their price to in some instances do things like fly in from another country to spend the time sitting watching what was going on having also reviewed the code and refreshed their knowledge in the preceding weeks or months.
Oddly enough I seem to remember several banks starting to (very slowly) update a lot of their really old* systems to more modern code and native hardware in the years after 2000, it was almost like something had happened to focus the minds of management and the accountants on the age and supportability of the systems their companies relied on.
*There were/are systems that are running basically the same code from the 1970's on emulators where they hadn't been replaced due to the complexity of the systems added to it and the risks involved, as as long as they could run the code faster on new hardware (even with an emulation layer to look like the original hardware) it wasn't deemed a major problem. IIRC one of the biggest outage in UK banking history came when a bank tried to split the computer systems of two of it's subsidiaries and had a booboo.
