I suppose if they are otherwise matched they would be ok, but that may be a tall order. Vbe should always be matched in parallel banks, the jury is still out on hFE matching. Matching within a batch is generally good enough, but I’ll measure to be sure in really high power applications, keeping the close grouped ones for paralleling and using outliers for drivers. Matching between units from different batches cannot be guaranteed, especially if years or decades apart. I’ve got 21193’s from 1995 to 2018 on hand, and I wouldn’t mix those except in emergency and with large emitter resistors.
It’s just that the non-G’s may be a decade older than the G’s because it’s been a while since non-G’s were made. Which makes it less likely to have good matching between the two batches. I wouldn’t sweat even 30% difference in hFE, but IMO 10mV difference in Vbe is too much for the typical 0.22 ohm emitter resistors. I try to get it to 3mV. Increasing the value let’s you tolerate more vbe mismatch. If you use say 1 ohm, you can get away with a lot, but no one wants that much wasted output voltage.
One thing's for certain: this part is a perforated emitter transistor with very good gain linearity and the manufacturing processor hasn't changed much, if at all. The real difference is in the packaging.