25A at 250V would be more than 8 horsepowers. A device drawing that kind of wattage is not something you want on your lap.

Try adding a circuit diagram. Also try to focus on the details actually relevant to your question.

@Kubahasn'tforgottenMonica That is a handwavy view. A digital signal at a specific sampling frequency uniquely represents a band-limited signal right up to the Nyquist frequency. A sine signal slightly below Nyquist frequency looks like a slowly beating signal at Nyquist frequency which is a superposition of the original signal and one slightly above Nyquist frequency. Since reconstruction removes the second signal, the original remains. We only need higher sampling frequencies because reconstruction (digital as oversampling and analog with filters) is less than perfect.

Semiconductors have a large temperature range with negative temperature coefficient for resistance, so even uniform transistors have a spontaneously developing hotspot problem when used in linear operation near their thermal limits ("secondary breakdown"). Of course the gate voltage variation of paralleled FETs is an additional big starting ramp for this kind of trouble.

@JonathanS. I'd have thought that the vast exemplar gate voltage variation would not bode well for linear designs. At least you'd have quite some difficulty providing circuit diagrams with dependable idling voltages... Also the specs for IRF540 state "IRF540 is a third generation power MOSFET with low on-resistance, fast switching and rugged design. It has a TO-220AB package, a 100 V drain-source breakdown voltage and a 28 A pulsed drain current rating." which sounds intended for D-class operation.