Your problem is the loop gain is higher for the LC components then the crystal so the oscillator is running at the LC combination and the crystal is just acting as a capacitor. (If you replace the crystal with a 6 pF cap you will get close to the 7 to 8 MHz oscillation you are now seeing).
LC oscillator, having a much lower Q will start up much quicker then a crystal oscillator in simulation.
Your inductors are very unrealistic for the 40 MHz frequency. They are appropriate for chokes at 40 MHz.
Multisim allows you to manually set the compution points in time. Sometimes this works better then allowing it to automatically select what it thinks is the necessary time increments between calculations. I usually always set compution increments manually when simulating oscillators. This is not your problem at present, it is the inductor values.
A 32 kHz crystal can take several seconds to startup in the real circuit so simulation is a very long time.
A 40 MHz crystal falls into the frequency zone that it could be fundament mode or third overtone mode crystal. I assume you are expecting it to be a fundamental mode. You need to give us some more info on the crystal spec. Is is fundamental mode for 40 MHz? Is is cut for series mode operation or parallel mode? If parallel mode what is its expected load capacitance for spec'd frequency?
If all you have is the component in software stock library look at the model. What is Co, Cm, Lm, and Rs in the model?
Attached is a colpitt osc for third overtone 40 MHz crystal. If crystal is truely a fundamental you can just eliminate (short out) the warp coil and eliminate resistor in parallel with crystal and run it as parallel mode. Circuit in this modified config will present about 10 pF of parallel load capacitance on crystal run as fundamental mode requiring 10 pF load on crystal to achieve stamped crystal frequency.