There’s two ways to blow up a 7135, overvoltage and overtemperature.
Overvoltage…
The AMC7135 is rated to 7V absolute max on both Vdd and OUT pins. Vdd gets fed with the Vcc voltage of the ATTiny, so provided the zener mod is properly done to keep the AVR’s Vcc at 5V or below, you’re good. You’ll probably blow up the AVR before you blow up the 7135.
Assuming everything else in the flashlight is zero resistance and the AMC7135 is off, the voltage on OUT is going to be the battery voltage minus the voltage drop of the LED at zero current. According to the MT-G2 datasheet (pages 7/8) the forward voltage drops at 0mA are ~5V and ~7.5V respectively.
Say you’ve got two cells (8.4V charged) and a 6V MT-G2, there’ll be 8.4-6 = 2.4V across the 7135 with the flashlight off. You’re OK. And if you’ve got three cells (12.6V) and a 9V MT-G2, there’ll be 12.6-7.5 = 5.1V across the 7135. You’re still OK.
But if you use 12.6V of batteries and a 6V MT-G2, there’ll be 12.6-5 = 7.6V across the 7135. Now what will 7.6V do? Probably nothing, but as you keep increasing the voltage on the OUT pin of the device the pin will start sinking current, increasing with increasing voltage across the 7135. This voltage/current combo will start to heat up the device.
If the 7135 has a FET output, this would be fine as long as the device doesn’t end up overheating. But I’ve got my suspicions the 7135 has a NPN transistor output, which means avalanche breakdown can happen, and that may not require much current (potentially just a few mA). Paralleling 7135s won’t save you here as different parts can have different breakdown voltages meaning breakdown current won’t be shared.
So 12.6V of batteries and a 6V MT-G2 might work - really it depends on the point when the 7135 breaks down. Personally I wouldn’t do it.
Overtemperature…
When the 7135s are on, they’ll dissipate heat. Again assuming zero resistance for the rest of the flashlight, the power dissipated in the 7135s will be LED current * (battery voltage - LED forward voltage at current).
With a 8.4V battery voltage and a 6V MT-G2 at 3A, the MT-G2 forward voltage curve says 6.2V. This gives (8.4-6.2)*3 = 6.6 watts to be dissipated in the 7135s. With a 12.8V battery voltage and a 9V MT-G2 at 2A (forward voltage 9.3V) you’re dissipating (12.8-9.3)*2 = 7 watts. But with a 12.8V battery voltage and a 6V MT-G2 at 3A (6.2V), you’re dissipating (12.8-6.2) = 19.8 watts!
How much power the driver can dissipate is proportional to the # of 7135s and how well the driver is heatsinked, but dissipating 19.8 watts is going to be a heck of a lot harder than 6.6 watts.
Overall…
If your driver fails, chances are the 7135s will fail short-circuit, turning your driver into a direct-drive driver. This will cause a very bright LED if you’re lucky, an open-circuit blown LED if you’re half lucky, or a shorted LED if you’re unlucky. The last situation shorts out your cells, in which case I hope you use protected cells.