At the speeds we're familiar with, basketballs and golf balls have elastic collisions. At orbital speeds, satellites are nearly inelastic. So fragment exit velocities lie between the two initial velocities, kv1 + (1-k)v2 for some k that depends on where each fragment came from. If they're colliding, the velocities must be somewhat different, so the weighted average speed has to be lower than orbital speed. So fragments usually don't survive many orbits.

I guess if a collision ruptures a pressurised tank, or causes an actual explosion then you could end up with a higher-than-orbit speed?

That's what I was thinking, Kessler syndrome should be impossible for objects in LEO since all debris orbits decay rapidly (probably 99.9% enter the atmosphere and burn up in minutes, the rest in hours)