Yes, there is a very formal and rigid system for validating the odds on all gaming machines.
In almost every regulated gaming jurisdiction (let's leave US Indian Casinos out of this for the time being), a third-party lab is hired to evaluate the code, look for backdoors, and verify the odds that are specified with the PAR (Probability and Accounting Report) of the machine being tested.
In the case of your roulette machine, that code was verified to make sure the random number generator created a valid distribution of numbers. If you had won a large enough jackpot (as those players did in your links), the gaming board technician would have been called in to verify the game's program code against a known verified version of the system as signed off by the third-party lab. When the checksums don't match...no jackpot. That's why every slot says MALFUNCTION VOIDS ALL PAYS.
That being said, there can still be bugs in the program code. When a machine is listed as paying a max of $2500 and a player wins $1,597,244.10, there's definitely a bug. Casinos, since they own the place, will err on their side and send you home without the jackpot. That's just the way it is. The manufacturer will get an earful and probably lose their positions on the casino floor until the bug is fixed.
Oh and look at your second link. The jackpot was "42 million". Wanna bet it was $42,949,672.95? (Hint: What is 4294967295 in hexadecimal?)
One thing I'm still not clear on. In the case of a machine that suggests it is simulating something physical, are the machines required to implement realistic odds or just accurate to what's official stated?
Do virtual card games have to maintain a proper virtual deck? Do virtual slots that suggest 20x20x20 reels have to implement 8000:1 odds?
In general, yes. When you're simulating a traditional physical gaming device (a pair of dice, deck of cards, roulette wheel, lotto balls, coin flip), it must match the real device.
From what I recall, card games are a very special animal in the eyes of the gaming board. When you simulate dealing from a deck, there must be a deck of 52 cards in memory. You cannot just pick random numbers from 1 to 52. I believe every hand must have a freshly shuffled "deck" as well.
The way video poker machines manipulate their odds is through the payout. Watch veteran players when they approach a machine. They'll note that some machines offer slightly different payouts for lower wins, and that's where the odds grind out. You can't change the probability of a full house, but you can change how much is paid back.
Slot reels are a loophole. Because you can't see the entire reel, how do you know it only has 20 stops? What if it has 20,000? This concept (known as the Telnaes patent, US #4448419) enabled virtual reels and much much larger payouts than the 8000:1 jackpots before this invention.
Later on, the invention of bonus games and etc kind of ripped up the virtual reel idea. But the math operates on this principle as far as the PAR sheets are concerned (i.e a certain set of stops on the reels starts a bonus game with another known set of odds).
Virtual slots definitely do not have to simulate the same reel size as their physical counterpart. I've only worked in online gambling, but almost ever slot game I saw had larger reelsets. Some don't even simulate physical reels, but instead just use a probabilistic model to decide what appears.
The Indian casinos are also highly regulated by the National Indian Gaming Commission. For some things, each tribe also has their own regulations that need to be handled, so it can be a major PITA work for or sell a product to more than one tribe's casinos.
If they can check the machine, they should have done so before allowing people to play. After all, they aren't paying back the losses on the "malfunctioning" machine.
It's to make sure the player or other third party hasn't messed with the code. And if the machine is paying out less than it should, there are penalties from the gaming commission.
If this was to prevent fraud, that would be what the signs on the machines say. "Malfunction voids all pays" is to minimize payouts, even when the gambler is not at fault, e.g. buggy software, non-verified software, unrelated tampering or tampering attempts. It's sufficient that "something" is wrong when the jackpot is won.
The argument I was trying to make was an ethical one: By accepting the gambler's money, the casino should take responsibility for the state of the machine at that point, thus incentivizing them to be diligent. To void a jackpot, they should have to prove fraud by the gambler or their associates to a court of law.
In almost every regulated gaming jurisdiction (let's leave US Indian Casinos out of this for the time being), a third-party lab is hired to evaluate the code, look for backdoors, and verify the odds that are specified with the PAR (Probability and Accounting Report) of the machine being tested.
In the case of your roulette machine, that code was verified to make sure the random number generator created a valid distribution of numbers. If you had won a large enough jackpot (as those players did in your links), the gaming board technician would have been called in to verify the game's program code against a known verified version of the system as signed off by the third-party lab. When the checksums don't match...no jackpot. That's why every slot says MALFUNCTION VOIDS ALL PAYS.
That being said, there can still be bugs in the program code. When a machine is listed as paying a max of $2500 and a player wins $1,597,244.10, there's definitely a bug. Casinos, since they own the place, will err on their side and send you home without the jackpot. That's just the way it is. The manufacturer will get an earful and probably lose their positions on the casino floor until the bug is fixed.
Oh and look at your second link. The jackpot was "42 million". Wanna bet it was $42,949,672.95? (Hint: What is 4294967295 in hexadecimal?)