I've seen many single-ended parts oscillate when inputs swing slowly, or are
biased near the switch-point. It seems the differential part would oscillate,
or do something undesirable, if both inputs were at the same voltage too long.
This would be the case if skew got too large.
If the diff_pair were terminated into a 3 resistor terminator then this probably
wouldn't happen since even if both drivers drove high, for example, the one
driving the R closest to VDD would be at a higher volatge.
The other thing that could happen, as might have been mentioned, is if the
diff_pair Zo is mainly due to Zdiff (tightly coupled+/- pair), then, if the skew
were larger than a risetime, the Zo could be 50% higher.