This thread is another one that suffers from vague terms. In my opinion, a game bombs when it sell serious less than expected. You can drool at first week sales if you want, but ultimately it's lifetime sales that matter. Zack&Wiki wasn't a bomb at 300k and Heavenly Sword probably isn't a bomb at 800k, but Final Fantasy XIII or Super Mario Galaxy at 2M would be the biggest bomb of all time.
GT4:P did 1.4M in just Japan and Europe, so it's reasonable to assume GT5:P will do somewhere between a million and two worldwide. Under a million would probably be a bomb, but that doesn't look too likely.
As for MGS4, it's hard to predict. The series has been in a constant decline sales-wise, but the previous part had some valid excuses (released around a lot of big titles, sequels on the same console usually sell worse etc.). The earlier titles were also on the PS1 and 2, which were vastly more succesful than the PS3. How many of the people who bought the previous games were devoted fans that will buy the sequel no matter what, and how many just got it because it was a great game on a console they owned?
This is clearly impossible to know beforehand, but I would say the amount of serious fans is greater than with your average shootan game, but still well under half of all purchases. I think anything under 2M would count as a bomb (not likely, but bombing means "doing worse than expected" doesn't it?), anything up to 4M as good as expected, and anything over that surprisingly good.
This is not the PS2 we're talking about here, and some of you people are throwing around some truly ridiculous numbers. The PS3 has a very small userbase that isn't buying a lot of games, and pretty much every sequel to a PS2 game so far has sold less than its predecessor. Surpassing MGS3 might be possible if some truly desperate bundling occurs, but reaching the first two parts is utterly ridiculous.