I've thought about doing this as well (Herms in HLT, while ramping HLT to strike) but I envisioned using a PID to control a pump. Place the probe at/in the pump outlet, pump into the TOP of the herms loop, and drain return to the tun. Once the outlet wort into the herms is back to the correct temp, the pid would turn off the pump, and the wort still in the herms would finish draining back to the mash tun.
(This, of course, assumes a gravity transfer system)
As the HLT rises to strike, the HERMS cycles would simply get shorter. And, in step mashes, you'd be in better position to step faster as your HLT has been heating up the entire time you've been sitting at a break.
Am I missing something key here? I could see how an insufficient flow rate could return a portion of your wort above your step, but returning it through a sparge arm should distribute that extra heat into your still-not-there-yet temp wort, and as a bonus help avoid thermal layering.