> reference it above ground as in certain parts of the Fairchild 670. I think in this case it is the valves that are in the regulator circuit that are referenced at 1/2 the HT voltage.
That's because the cathode of the power tube is up around +300V, and the volt-amp tube not much below (maybe 150V?). So the only safe place for the heaters is up in that area, 150V-300V, so neither tube exceeds its heater-cathode voltage rating (100V-200V).
Same thing can apply with cathode followers, which often have cathodes up over 100V. If the heater insulation is only good for 100V, then the heater winding has to float well above common ground. If you want to try to run cathode-followers and semi-grounded-cathode stages on the same heater winding, split the difference: 50V-100V on the heater winding.
There is also a fashion for not having a solid voltage reference on the heater winding. Theory is that heater leakage will make the winding find its own voltage mid-way between all heaters, and if one heater shorts to cathode the system will keep working. This seems fishy to me: it works when it works, but can fail in odd ways. Also you almost always have to hang a 0.22uFd or so to ground to supress audio and RF leakage wandering around the heater circuits, so a heater-cathode short probably will sicken the amplifier.
When all your cathodes are semi-grounded: grounded heaters will work, and if the heater insulation is perfect they work very well. But 0.1% of tubes will have small insulation leakage. If the heater is held positive of cathode, the resulting hum is less than with a grounded or negative heater winding. So many mass-produced amps bias the heater up around +50V. In most cases, grounding makes no difference, but once in a while you find a tube that buzzes with grounded heater but is clean with +50V bias. Obviously in mass production it is better to add $0.12 of parts to every amp than to test every amp and tube-roll the occasional leaker.