What I suggested was that you consider generating a free-running crystal oscillator using a hex inverter. The output from one side can be taken from pin 8 as I suggested, which makes it pin-compatible with the older-style crystal oscillators built on this package format. The fact is that there are three additional inverters remaining in the same package, and you can use those similarly to the ones on the "high" side of the package, which yields a buffered oscillator, the output of which is then on pin 6.

I further suggested that you examine and consider a circuit that's available on XILINX' website in an app-note, which enables you to switch between two clocks at entirely unrelated frequencies in a manner that's free of glitches or runt pulses.

It wasn't clear that you needed such a circuit as the second one to which I referred, but in the event that you needed to switch dynamically, it would be very useful.

Yes, I believe you'd benefit from avoiding the use of the built-in crystal oscillator on the MCU in question because it's difficult to predict precisely how such an oscillator will behave when the components are switched in any way.

RE