Even with the splits in the grid, the time lag is still significant, I would think. The western interconnection appears to be 3,700,000 from the corner of Arizona to the corner of British Columbia. The wikipedia article doesn't address the time lag (except perhaps implicitly).
Indian Point to Yankee Nuclear is 230 km, IP to Niagara Falls is 440, and NF to YN is 500 km. At c (electronic energy propagates slighly slower), 1 km = 4.32' (arcminute) of lag. That means the 'legs' of this triangle have lag of 16, 32, and 36 degrees. That around half a Hz of frequency out of phase, or ~0.9%. By comparison, the 2011 tornado outbreak caused 0.09 Hz deviation.
So there must be some accounting for this propagation error, either as losses, or engineering.
What I suspect (totally not my field) is that the high power interconnects between large generators do have this triangular lag (the incoming lines lag from the plant's reference frame), but they use quadrature boosting or similar to match that particular interconnect.
T&D (transmission & distribution) result in ~6% of the net energy being wasted. I imagine phase lag "feels" like impedance, with the real component acting as resistance.
Also, it's worth noting part of the appeal of HVDC, aside from line reactance and skin effect, is you get to choose your output frequency.
