Agreed. Logic simulations produce and propogate generic X's, not real circuitry.
This is not always true. Depending on the logic family and actual circuit design of the flip-flop, it may be possible for the output voltage to momentarily sit in the guard band between the voltages that are defined as a valid "0" or "1" according to the design spec of the the logic family. If you don't wait for the feedback of this flip-flop to resolve to a voltage well outside of the guard band, then it is possible for two different gates connected to this flip-flop output to make (or not make!) opposite decisions as to whether this signal is a "1" or a "0". Not good if you are trying to, say, clear a counter and only some counter bits clear, or transition to a state machine NextState and end up in the wrong state, or even an illegal unrecoverable state.
There are also other problems that propogating a metastable signal might cause, including increased shoot-thru power consumption, ground/power bounce and thus signal integrity issues with the local surrounding logic, etc.
IMHO. YMMV.