In this case, the context would be:
"Holy shit guys, NK just launched a rocket that looks like it's headed for LA!"
[3 minutes later]
"And now the US just launched 14 rockets, and they might be headed for us, but they're probably headed for NK."
China's best option is to wait and see, because if they decide to fire, and were wrong about China being targeted, they'll see another round of launches three minutes after that one, and those will be headed for China.
They've played this all out. MAD really does work.
China doesn't want to enter WWIII any more than the US does. And the US response to an NK attack is very predictable.
Any strike on NK has to decapitate it instantly, or the conventional artillery along the border will pound Seoul into rubble in the meantime. Ballpark figures are 10,000 pieces, some dug in, some mobile, each capable of firing 10kg high explosive and keeping it coming until their ammunition runs out. It's a hard problem for any military planner.
If the SK capital was on their south coast and it was only farmland within artillery range of the border than NK would be an easy problem to solve (obviously if that were the case NK would have developed other weapons to compensate). Notwithstanding China of course.
You would also need to somehow move your DMZ-overrunning force into position without that manoeuvre itself triggering the artillery bombardment.
A ballistic launch towards NK looks an awful lot like a ballistic launch towards China, is the problem there. Will China "wait and see" where it lands?