James,
You may consider that if you are using just a crystal itself that there is usually not a good justification for trying to share them between two microcontrollers. Here is a short list of reasons why.

1) The cost of one versus two crystals is very small in consideration of the total cost of an electronics assembly.

2) When a crystal is attached to a microcontroller, with its attendant capacitors, the trace lengths need to be kept VERY short. If you try to connect to two processors it is almost for sure that the length of the traces would be longer than what is generally recommended for a stable oscillation frequency.

3) If you do use a single crystal on one microcontroller it is then necessary to buffer the X2 signal from the first microcontroller via a very close by high impedance CMOS gate part. The output of that gate can then drive the X1 input of the further away second microcontroller. Modern TinyLogic single gate chips are a natural for this type of buffering...but do consider that in a low volume project such as yours that the cost of this extra gate part is nearly the same as a second crystal.

So in conclusion, my recommendation is to use two separate crystals OR to use an off the shelf oscillator can that has a buffered output.

Michael Karas