They never actually used the stolen tech, right? It wasn't good enough. They hit the pedestrian solely on their own tech, which (clearly, in hindsight) wasn't good enough either.
It wasn't disclosed on whether stolen IP was used - Uber had (still have?) to cooperate with Waymo, inspecting their hardware and software stack.
The pedestrian ultimately got hit because a lot of fallback systems failed: They disabled the software's ability to emergency-break (1), they disabled the cars' systems to emergency break, and the operator of the vehicle was distracted, not reacting in time.
(1) This was actually done due to false-positives, i.e. the car slamming on the breaks in non-emergency cases. (I find that's a relevant point, because without that detail it just sounds more ludicrous than it already is).
The IP wasn't the problem in the accident, rather the implementation. Simplified, the software was tuned in favor of ignoring obstacles as false positives. By the time the system acknowledged the immediate danger it was too late. No tech, no matter how good the IP, could stand up to being tweaked like this.
And the emergency breaking relied on the human operator but the system was not designed to warn the operator.
