Definitely the different setups require different rest periods. Might be a good practice to try a long rest with progressively shorter test periods. That's if you don't mind blowing one lamp. Otherwise, measuring the volts of your pack will give you a good idea what is safe. Here are measurements taken from a Mag85 6C using 9 Gold Peak 2200 mAh 4/5 Sub C batteries:
12.86V on batteries only, no load with one hour of rest
12.29V at the tailcap, no load with two hours of rest.
11.79V at the tailcap, no load with two hours of rest and 1 minute of use.
11.49V at the tailcap, no load with two hours of rest and 5 minutes of use.
11.20V
at the lamp, under load with about 7 minutes of use.
10.99V at the lamp, under load a few seconds after start up from the above condition.
10.86V at the lamp, under load from less than 1 minute after the above condition until 4 more minutes into the run.
Resistance and current are players. At the lamp, under load with 2 hours of rest and about 8 minutes of use:
3.41A at start up that drops quickly to 3.37A then eases down to 3.31A and stayed there for several minutes until I got board watching the meter.
Resistance and current aside, it's possible that KingSmono's suggestion of 12.6V is a safe and reasonable benchmark for a no load measurement for a rested NiMh power source for the Mag85 HotWire. Of course, YMMV.
TKO -
We don't often go the speed limit in this here burning tungsten forum.