M4M4X’s review is worth a look for anyone thinking about buying one of these, or having problems.
See video review: Fireflies PL47 and E07 // some more vids added
For the want of an extra few mm of metal, or a little research into cell dimensions and thinking about tolerances when designing the metalwork, this project seems to have been a fail for some.
As for the electronic/firmware design and integration, well it’s early days yet to know how durable the LEDs will prove to be.
Fundamentally a torch that (unmodified) can cook itself in a minute or less, along with any other failings, is a bad design. It’s not the user’s fault, nor the fault of the dumb cell (that may or may not even fit), it’s just lack of design, integration or development testing, never-mind independent review. Also entirely predictable, given the hardware and firmware capabilities and limitations, when constructing such an over-powered device with so little thermal capability.
Between the manufacturer, the electronics and the firmware consultants, a way should have been found of controlling this beast. And I don’t mean just shipping it in muggle mode with temperature control set to an unusably low level. Still, it comes with a multi year warranty, so no need to worry 
Any such ideas would probably also be usefully applied to e.g. the FW3A, though at least that one should be able to mostly run purely on 7135s, with the FET kept only for special (short) occasions.
What is certain is that changing LEDs, should it be required, will not be easy, this torch seems to be purposely designed to frustrate hardware modders.
With these small, hot, thermally challenged, multi-emitter torches, with the output power of a Q8,
Edit, or nearly twice that, in the maddest one. Seven emitters WTF? What could possibly go wrong ? does this seem like a plan ?
Just because you can make it, and persuade some people (noted) to prostitute themselves to try to make it work, after a fashion, does not mean that it it is anything that I desire. Absolutely the opposite.
Thermal control is probably the most important challenge for hardware and firmware designers. And frankly I don’t think the way it is being approached at the moment has much merit or future. Nevermind requiring the user to fiddle about with it themselves.
This is where the effort should be being applied, rather than just titivating switch and aux. LEDs, pretty though some may think them.