I've recently fooled around with various 7404-pin-compatible IC's in crystal oscillator configurations, and found that in the range between 8 MHz and 25 MHz, it seems to matter very little whether the inverters are schmitt-triggered or not. Now this is a relatively small sample, but what I've used is a 1K resistor in feedback between pins 12 and 13, and between 11 and 10, with a 150 pf cap between pins 12 and 11. The same is possible on the "other" side of the package, if you need two oscillators, as I did. The third inverter, between pins 9 and 8, or between pins 5 and 6 on the other side, is used as a buffer. This is soldered to the test fixture and the various IC's are substituted via a socket. The crystals are socketed as well, and, while this is a relatively uncommon set of connections in the electrical sense, the circuit seems to work with AC, ACT, AHCT, HC, HCT, FCT (didn't have any FC), AS, ALS, LS, 74S and standard 7404 as well as all the 74xx14 types. I didn't try 74C, and did experience some peculiarity, (startup trouble) using 74L types. It works with fundamental mode crystals up to 40 Mhz by which point not all the technolgies work equally well, probably because they'd like different feedback resistors.

I've used this scheme on a number of CPLD and FPGA boards that use an external oscillator, particularly those that required two crystal oscillators, with essentially complete success. I've also done this with 0805-packaged passives soldered directly to DIP IC pins, too.

I'd expect some challenges at the extremes of the frequency ranges, but, having had success with frequencies as low as 2.4576 MHz and as high as 40 MHz, I wouldn't be afraid to use a schmitt inverter if that's what I needed elsewhere in the circuit.

My guess would be that (1) an external oscillator driving both MCU's would work well, and that (2) if you MUST, for some reason, use a crystal and not an external oscillator, recognizing that that one pin (18) is the input and the other (19) is the output might be a clue as to how to share the crystal. One MCU would be the oscillator and the other would be driven via a buffer as though it were an external oscillator. It will require some fussing with the biasing and buffering, however, and IMHO, it's not worth the trouble.

RE