Two directions.

Let's examine a slice of the booster. Going vertically you have one segment, then the joint, then the next segment. The O-rings were in that joint and had some ability to move horizontally.

As designed the joint would always be in compression, the O-rings sandwiched between two big pieces of metal. If they moved horizontally in the space they had it made no difference, their job was simply to keep the 1000psi inside the booster inside it. Going inward there was a layer of putty that could stand up to the heat but was useless for sealing.

Unfortunately, when the engines lit the whole booster stack twanged a few inches. A joint meant to always be in compression was suddenly for a moment in tension--the two pieces of metal moved slightly apart--gas could now go above/below the ring. If the rings were pliable enough they got slammed against the outside of their groove where the pressure against the joint stopped the escape of gas--examination of the boosters showed blow-by but it cut off soon enough that the mass of metal was enough to absorb enough heat to avoid catastrophe.

But that night was very cold. And it was very calm--the boil-off from the LOX tank was simply dumped overboard and the booster that failed was downwind. The point of maximum chilling was between the booster and the tank, the lowest segment joint got the worst of it. And that's where it failed.

When the stack twanged the ring didn't slam against the outside quite fast enough--some exhaust leaked past and tore up the ring. But the gas still had to go out the joint--and the shuttle fuel used aluminum. The ring wasn't sealing the joint but enough aluminum solidified out against the still-cold metal of the joint that it sealed the gap and Challenger roared into the sky. But as it went faster and faster the vibrations grew stronger--and eventually the really sloppy weld let go. Even that didn't doom the mission, there was enough fuel to tolerate the pressure loss. But the leak was pointing at a strut and the tank with a whole bunch of LH2 in it. Neither was designed to stand up to that.

There was also a second failure that got little attention: the putty. As intended, it should have covered the entire gap, the force would have been evenly applied and it probably would have made it. But the putty was spread and the segments placed together--in atmosphere. Air was trapped and compressed--and the putty gave way letting it out. What had been an even layer now had holes in wherever the weakest spots were--and that concentrated the escaping gas from the booster. And why wasn't that caught? Because in the static testing someone had gone inside and made sure the putty job was good. Easy enough in a booster laying on it's side, but the Shuttle was stacked vertically.