It can happen that a bad solder connection on thermal pad kills your MOSFET
basically the SIR800 is thermally at its limit on the driver PCB, bridging springs and putting fresh cells in a hot light can get it to thermall runaway
There are basically 2 factors that generate heat in a MOSFET, conduction and switching losses
1. conduction losses, as simple to understand as a resistor
BSC009NE2LS5IATMA1
lateste Infineon MOSFETS with Optimos-5 technology have 1-1.5mOhms with relative low switching losses (20nC@4.5V)
https://www.mouser.de/ProductDetail/Infineon-Technologies/BSC009NE2LS5IATMA1 (look out for the excact type there are differences in Gate charge)
the suggested from MikeC I am and Intl-Outdoor are using for quite a while has 1.2mOhm at 3.5V gate voltage at 25° Tj
so as they do n ot stay cool lets add about 40% to get to Tj=125°C worst case
So 1.68mOhm@30A= 1.5W thermal loss
SIR800DP
https://www.mouser.de/ProductDetail/Vishay-Siliconix/SIR800DP-T1-GE3?qs=sGAEpiMZZMshyDBzk1%2FWi0mX49X1Ojk8bkorZS29FXY%3D
this one is nice and gives a graph for cool and hot junction resistance
3mOhm at 3.5V = 2.7W
I had recently working on a driver with a chineese company fake Attiny85s from a big chineese distributor,
maybe same problem with the MOSFET if HL is trying to source cheaper locally
on Turbo we can ignore switching losses, but they can add at very high PWM cycle
2. switching losses
to expalin simple between on and off the MOSFET transits slowly from low resistance to very high
in that short time the increased voltage drop creates heat
There are various models to calculate those switching losses, some more simple use the Gate charge and voltage to estimate a rough value
For Synchronous buck or boost converters some companies support more sophisticated formulas or even programs that calculate it (On Semiconductor, TI, Infineon, Nexperia)
to keep it simple the faster it switches and the less “dyanic reesistance” the gate has the fewer losses
also the Gate resistance ghas a role, the smaller the better
BSC009NE2LS5IATMA1
rise time 6ns
fall time 4ns
Gate charge 20nC
Gate resistance 1Ohm
0,53W switching losses at 22kHz, even at 99% PWM cycle
SIR800DP
rise time 10ns
fall time 8ns
Gate charge 41nC
Gate resistance 1.2Ohm
0,9W switching losses, at 22kHz, even at 99% PWM cycle
Overall losses when ramped to 95%
BSC009NE2LS5IATMA1 1,87W
SIR800DP 3.34W