Is there a good list of "worst software project failures ever" in the history of software development?
For example in Canada a " gun registry " project spent around two billion dollars.
This is of course, insane, even if the final product "sort of worked".
I have heard of an FBI Case file system which there have been several attempts to rewrite, all of them so far, failures.
There is a book on the subject (Software Runaways). There doesn't seem to be be a software "boondoggle" list or "fiasco" list on Wikipedia that I can see.
(Update: Based on a response of human sympathy Therac-25 would be the 'winner' of this question, except that I was internally thinking more of Software projects that had as their deliverable, mainly software, as opposed to firmware projects like Therac-25, where the hardware and firmware together are capable of killing people, and also, the question was more intended to address boondoggles and bureaucratic failures rather than tragic explosions, or deaths. In terms of pure software monetary debacles, which was my intended question, there are several contenders, and an interesting future community wiki, is "what are the common traits among large software project failures with budgets over $100 million US".)
The Ariane 5 integer overflow  :
This incident gave rise to two interesting articles that discuss the benefits/drawbacks of using design by contract  techniques for constructing reliable software:
Let us all remember that our carelessness, no matter the profession we choose, has real consequences. http://en.wikipedia.org/wiki/Therac-25
Duke Nukem Forever . http://en.wikipedia.org/wiki/Duke_Nukem_Forever
Mars Climate Orbiter  23 September 1999 Orbiter Crash landed on surface due to metric-imperial mix-up http://en.wikipedia.org/wiki/Mars_Climate_Orbiter
Netscape 6.0. It cost them a considerable lead in the browser wars.
Here's Why: "Things You Should Never Do"  http://www.joelonsoftware.com/articles/fog0000000069.html
The worst one is when you are responsible for it.
I recommend you take a look at thedailywtf.com  for numerous articles about software failure. In many cases the stories are true; only the names have been changed to protect the guilty. http://thedailywtf.com/
The worst?... How about the Ballistic Missile Early Warning System (BMEWS) nearly causing global thermonuclear destruction after detecting the moon rising over the horizon and erroneously classifying it as an incoming missile attack from Siberia ( Chapter 2 "Boardwalks across the Tar Pit" from Mechanizing Proof by Donald MacKenzie ).
That's pretty scary $h!t! http://books.google.com/books?id=QiMS8t4V%5F0cC&printsec=frontcover&dq=Mechanizing%2BProof
The Netscape rewrite.
In Britain, the NHS National Programme for IT . It might reach GBP 20 billion, so 30 billion USD.
It's now part of law that every British contractor must work on the project at some point... http://en.wikipedia.org/wiki/NHS%5FNational%5FProgramme%5Ffor%5FIT#Costs
One case, more of a classic project failure, closer to the Canadian example mentioned by OP:
AKE  (Finnish Vehicle Administration, a public agency) ordered a project to overhaul its information systems in 1999, mainly from the companies TietoEnator  (now Tieto) and WM-data  (now part of Logica).
It was supposed to be ready 2003, but has been continually postponed. In 2007 it was estimated that it'd be ready 2011. So right now the project has been going on for a decade, and it will be at least 8 years late!
The budget has gone through the roof too: original estimate was 16 million €; actual total costs as of 2009 have been more than 50 million € (~70 million USD), more than 300% cost overrun, so far.
From news articles, it seems like a proper mess of every kind of leadership, coordination and requirements problem: teams with overlapping responsibilities; no-one having an adequate picture of what was actually needed when the project started; the project responsible at AKE having been changed at least 5 times.
So, nothing as spectacular as Mars landers crashing, or people dying because of this (afaik!), but these sort of things are probably among the most common failures in this field. The main consequences: loads of wasted taxpayer money, and screwed reputation for the software/IT industry. :-\
Sources (in Finnish):
How about the Pentium bug- http://en.wikipedia.org/wiki/Pentium_FDIV_bug  http://en.wikipedia.org/wiki/Pentium%5FFDIV%5Fbug
Three people died (and three others horribly injured) due to a small and rare race condition in a medical radiation device. 
A really sad story.  http://en.wikipedia.org/wiki/Therac-25
Windows Vista comes to mind
Chrysler's infamous Payroll application, that spawned the Extreme programming concept.
Something like 5+ years of development, and it never cut a single check, then had the plug pulled. Chrysler then banned the practice of Extreme Programming.
Here's a list of some of the worst:
These include the Morris worm, the Kerberos vulnerability, the Therac-25, and the Mariner I space probe failure, among others.
According to The Mythical Man Month , OS/360. http://en.wikipedia.org/wiki/The%5FMythical%5FMan-Month
How about the Sergeant York gun? http://en.wikipedia.org/wiki/Sergeant_York_Gun
"Unable to hit drones moving even in a straight line, the tests were later relaxed to hovering ones. The radar proved unable to lock even to this target, as the return was too small. The testers then started adding radar reflectors to the drone to address this "problem", eventually having to add four. Easterbrook, still covering the ongoing debacle, described this as being similar to demonstrating the abilities of a bloodhound by having it find a man standing alone in the middle of an empty parking lot, covered with steaks. The system now tracked the drone, and after firing a lengthy burst of shells the drone was knocked off target."
What about the Denver Airport Automated Baggage Handling System? The system's budget was $193 Million dollars of 1994 - and software delays were costing $1 Million dollars a day. According to Scientific American (September 1994), the system consisted of 100 networked computers connected to 5,000 electric eyes, 400 radio receivers and 56 bar-code scanners.
The Distributed Computing Environment, at least for OS/2.
The Taligent operating system.
The IBM Workplace OS.
Large chunks of money ($2 billion for the last, according to Wikipedia), large chunks of time, and you have probably never heard of any of them.
But those are just the ones from my resume.
The 1985 failure by the IRS  to adequately test the new (but overdue and over budget) Sperry Univac system.
The agency sent refunds of tens of thousands of dollars to people who were owed nothing. They failed to send refunds to people who were owed money.
The departments were so backed up that ceiling tiles were removed so that tax returns could be shoved up into the ceiling. Returns (including checks from taxpayers) were flushed down toilets and taken home by employees to be thrown away. All so employees could look like they were keeping up.
In the wake of the fiasco, a new $20 billion plan was put into place to overhaul the system. It was also a fiasco. http://books.google.com/books?id=YtzFUqrQu1IC&dq=the%2Bgreat%2Bamerican%2Btax%2Bdodge&printsec=frontcover&source=bn&hl=en&ei=EIBCSsS%5FDILGsQOwraDMDw&sa=X&oi=book%5Fresult&ct=result&resnum=4
Air Traffic Control System. The US still uses a card-based system designed in the 60's. I know that at least 1 major program to computerize it failed. If the system were computerized correctly, airport congestion would be mitigated
Definitely not the worst failure, but probably one of the most stupid...
Another payroll I don't think has been mentioned is Wisconsin's payroll system. it has cost $28.4 million so far but they reckon $12 million more is needed.
This HAS to be a software failure imho.
At least Therac-25 shipped.
My vote goes to the California Department of Motor Vehicles, which hired Tandem and Ernst & Young to to replace its aging IBM mainframe-based driver registry system with a shiny new one. The five-year project, launched in 1987 with a budget of $25 million, was cancelled in 1994 after $44 million had been spent and there was still no delivery date in sight. The state then spent another half million dollars to find out what had gone wrong.
Waste Management suing SAP for $100 million  seems like it could be on the list in terms of dollar values on a failed project. http://www.itworld.com/waste-management-sues-sap-080327
Duke Nukem Forever...
Warren, did you look at Coding Horror article list? the long dismal history of software project failure  Quoting his quote from spectrum.ieee's why software fail article  In October of 2005, "the giant British food retailer J Sainsbury PLC had to write off its US $526 million investment in an automated supply-chain management system. It seems that merchandise was stuck in the company's depots and warehouses and was not getting through to many of its stores. Sainsbury was forced to hire about 3000 additional clerks to stock its shelves manually" http://www.codinghorror.com/blog/2006/05/the-long-dismal-history-of-software-project-failure.html
Microsoft server crash nearly causes 800-plane pile-up http://www.techworld.com/opsys/news/index.cfm?newsid=2275
Goldmine CRM. Without doubt.
Y2K - Have anyone actually calculated how much it costs all over the world to fix this?
Seriously - there is no excuse for that crap. http://www.uca.edu
Samsung PC Studio... Samsung New PC studio...
You can say what ever you want but this is the fact... :)
Winamp 3. (Nullsoft media player)
Can I refer you to the websites of the UK government?
The SABRE rewrite. I'll have to search for the name.
There is a company in Boise, whose name I will not mention. Their purpose was to build a shopping engine that would be used for female shoppers to enjoy a virtual-mall-like shopping experience. Instead, the project had 68 high-end developers (enough to build an operating system), and millions of dollars hemorrhaging each year, and the withered away over 2 years time.
Munich migration to Linux. (It counts as a software project, because they are writing their own Linux distro.)