Smajdalf:
I have little experience with such "high" power but from theory: lot of current needs large FET -> large FET has high gate capacitance -> with "weak" Arduino pin (made even weaker with 1k resistor you suggest) it will take "long time" to turn the FET on or off -> power dissipation grows. But this may be issue only for even larger FETs? I don't know.
Mmm... It looks like you're right, never tried to work with those very high current FETs. I didn't look at that parameter, also wasn't aware of it going up much with increased peak current. It kinda makes sense: bigger current will mean bigger chunk of silicon for the gate, and that means more charge needed to charge it up.
The SI7858BDP-T1-GE3 as I suggested (selected with Mouser's selection function) has a gate capacitance of 5760pF. Definitely needs a resistor on the gate at that kind of capacitance - 250Ω for 20 mA, or 270Ω as first higher R24 value - otherwise fair chance the port will fail after some time doing PWM. Switching on/off now and then it may survive but still not a good idea.
Now the good thing with these specific MOSFETs is that 1.4V Vgs means the Rds(on) is down to about 6mΩ, and that voltage is reached in 0.53µs already (based on 270Ω and 5V signal voltage). The first 0.4V (minimum Vgs(th), maximum is 1V) it doesn't conduct at all, that's the first 0.13µs, so about 0.4µs this MOSFET is in partial-on state when switching on.
The switching off part takes a bit longer. To switch fully off (drop the gate to 0.4V) takes 4.06µs. The first part, until Vgs = 1.4V, it's fully on, to drop from 5V to 1.4V takes 2.05µs, so that's 2.01µs in partial-on state when switching off, or a total of 2.4µs per PWM cycle.
Of note is also that the switching takes 4.6µs per cycle, that may affect the PWM duty cycle at high frequencies. At 1000 Hz it's just 0.46%, probably not noticeable. At that frequency the MOSFET is 0.24% of the time in partial-on mode generating extra heat (which indeed can add up quickly at 30A), a number that can be reduced by reducing PWM frequencies.
That's getting a bit harder to calculate, not just because there's no data on the actual Rds values, but also because very soon the Rds will start limiting the current itself. If it would be at 0.4Ω the current goes down to 15A but power dissipation is 90W - if that's the whole switching period it'd be 0.22W total, while the constant on current would mean a dissipation of about 2W. So 90% or less duty cycle at 1000Hz should reduce the overall dissipation.
The MOSFET still needs a pull-down resistor (missing from the latest schematic), 10k will do just fine for that as all it has to do is keep the thing switched off during startup. When doing PWM the port will pull down the gate as fast as it pulls it up.
That was a very interesting note! Hope I understood everything correctly, if anything is wrong in this post, please correct me!