In addition to other events, Sun entered dotcom boom a bit from a wounded position due to high volume of recalls caused by radioactive cache in what was to be flagship processor. Meanwhile IBM had a lot more revenue streams (including still x86 machines on low end, AS/400 on middle size, and Mainframe and supercomputing business at the top) that weren't much impacted by fickle nature of dotcom.
Static RAM they bought from IBM; they trusted it too much, didn't include ECC in its path to CPUs, and completely torched any chance at becoming an enterprise vendor by blaming their customers, making them sign NDAs before Sun would "help" etc. You really don't want customers to describe an "enterprise" experience as "They treated the whole thing like a cover-up"https://www.computerworld.com/article/2596346/more-users-sla...
That and what PeterStuer's describes as "They died because they were unable to rapidly change their businessmodel that relied on an expensive sales staff vendoring relatively big ticket items." killed them dead.
Specifically startups weren't able to buy Sun's high quality x86 systems in medium quantities from Sun's designated channels, there was a fatal gap between what you could put on a credit card and placing million dollar enterprise quantity orders. So startups bought from Dell which was actually willing trade money for hardware, learned how to deal with those cheaper systems' quirks, and if successful by the time they were big enough to buy Sun systems in enterprise quantities it was way too late. This of course before the cloud became a thing.
Something like what's discussed in the recent topic "Bugs that cost money" https://news.ycombinator.com/item?id=34103323 If you won't accept money from the people who want to pay you it for your stuff you aren't likely to survive.
I think I remember the problem was something like bit flips due to alpha decay from trace material in the chips’ packaging. Not physically dangerous but hard to diagnose and fix.
