this would normally be a dire situation for any high volume, high profile product launch and we would hope good PM'ing would prevent this sort of thing. But for sake of the interview question, we have to accept that stuff does happen. First get clear articulation of what has failed. A clear problem statement. A cryptic mail from the factory prepping the line for launch "omg, the touchscreens from units on line 2 aren't working" isn't going to cut it. Form a team - typically pulled from that products PDT that your running - the HW, SW, FW, Ops, Q&R, factory rep and start brainstorming and reviewing the data. Where was the failure found? in Validation - in a hardware or software lab? at a customer site during their beta testing? during volume production line prove out... if it's a software bug, get it entered in the bug management system if there is one so you can the brightest minds working on it globally. Filing a bug formally forces definition of the failure, setup so others can root cause it, etc. did a hardware component fail? is it random? It would be hard to imagine someone picked one lemon component supplier and it systemically fails at launch - such a bad part would have failed much earlier in tests. If it's randomly failing, get clarity on the conditions under which it fails - certain power conditions? temp/humidity conditions? is some aspect of the electrical design on the edge of one of the IEEE specs and tolerance stack ups have occurred. Schedule is king usually. So is cost. To protect schedule, do we launch anyway? again it goes back to the nature of the failure. Do you launch and jeopardize the valuable brand of your employer? The salient point is - form a team, gather data, understand the failure, and forumulate recommendations for management. All quickly. I wouldn't want to make an uniformed decision, and give direction to launch anyway without advising everyone that will affected later if it blows up post launch of all the risks.