Your observations are very good and those areas should be looked at a bit nmore. The uneven support of the deck should be reviewed and fixed if possible.
The welds in the stained areas deserve further examination, perhaps VB at 8-10 psig and/or PT. Do the stains contain product residue? Are they near legs? Sometimes cracks close up when the deck is flexed from landing but they open when the deck is flexed from floating. If the product has an ethanol content you might be seeing the start of some ethanol induced stress corrosion cracking. Does the product side of those seams show anything unusual like excessive buildup of rust in the lap or a leg carrying an unusually high load? In the picture of the leg lifted off the floor, it is not clear if the lap behind the leg is within 12 inches of the leg but it does not appear to be welded as it should be if it is that close.
A steel double deck is a heavy structure for a floating roof, while it should be fairly flat there will be some sag in the deck plates and what the pictures show seems not too out of the ordinary. They also are somewhat flexible in long spans but can be very stiff locally. A drawing would show the leg size, quantity and placement; each often supports somewhere around 400 square feet.
It is unusual in my experience for small temperature changes to cause large deflections but I'm not overly surprised at what you are seeing. My guess is that nothing is terribly wrong but the deck is flexing in response to uneven support conditions and the temperature change just happens to be enough to trigger a movement where it can most easily occur. There will be one or more legs in that area that are carying a higher load, and may be subject to premature failure.
It is possible that this deck was either not built flat, or due to floor settlement is not being supported flat (or both). If this is the case then repeated flexing from leg supported to floating may cause more of the type of leak you are seeing. I'd also look to see if any microbial induced corrosion or ethanol stress corrosion is present.
The highly loaded legs may fail sooner than the others especially if they are not perfectly vertical, which changes each time the deck lands. The higher stress in the deck at leg(s) carrying higher loads may lead to leaks in those locations.
A review of the leg corrosion, loading and buckling strength may be instructive. An elevation survey of the deck and floor at the legs would tell you which ones are long or short and where the floor is not in the conical plane of an ideal bottom.
Before you put this deck back in service, I'd want to resolve the mystery of the stains, look at the legs and their connections where the legs are carrying more load than others and make the deck flat when landed.
