I've never put the ERF2030+ in an N2/98VHP amp unit, but I have installed them into a Connex 4400 Turbo amp unit which is just half the size. It uses 4 FETs versus 8. I can tell you one thing that is absolutely a fact; they are NOT just a drop in replacement. The ERF2030+ FETs have a much higher gate capacitance, thus a lower input impedance. Just swapping the FETs without changing anything else will net you less power before you get into unacceptable distortion issues. The 4400 Turbo uses 2 turns on the input transformer, so 4:1 impedance radio, I had to add a 3rd turn to make it 9:1 which cleared up most of the distortion (non-linearity) I was getting. The rest was cleared up by tweaking the capacitors across the transformers. I can't remember if I changed anything else or not. The bias should be set by the drain current, not the gate voltage (this applies to all FETs though if you care about class of operation). I set my amp at about 50 to 60mA/FET, so 200 to 240mA total, with no RF input signal. I was able to get 100W PEP in AM with a standard 4:1 peak/carrier ratio, so 25W dead carrier, without any heating issues. The IMD3 was a bit more than I care to emit, so I backed it down to 20W/80W-PEP and it was clean as a whistle with input from my bench signal generator. I think it took about 4.5W-PEP drive to get the 80W-PEP output.
Now, having said this, a matched set of 4, GENERIC branded, IRF520s made 20W/80W-PEP without heating problems, and 15-17.5/60-70W-PEP with my obsessive distortion limitations. The factory, unmatched, IRF520s did 15W/60W-PEP with "ok" distortion levels, but 1 FET ran significantly hotter than the other 3. For my IRF520 experiments 50mA/FET bias gave me my best results with the generic ones that I had. Since the factory set of 4 were unmatched, I'm unsure of how equally the 200mA total was divided between them. Obviously the hot one was getting more than it's fair share.
My conclusion would be, at least for the 4 Mosfet amp, you'd be much better served by sticking in a tightly matched set of name brand (not the random generics that I had) IRF520s, properly setting the bias via drain current, and not overdriving them. Trying to squeeze out every ounce of power seems to be the cause of most of the FET failures, usually combined with distortion from the radio, improper biasing, and unmatched devices, these non-RF FETs don't stand a chance. Truthfully, the extra 10 to 20W-PEP that I get from the 2030+ FETs, doesn't really travel any perceptible distance further than without it. Sure, I can see it on the meter, but it's practically unnoticeable on the receiving person's end. Swapping in a matched set of IRF520s and biasing properly is much simpler than changing to the 2030+ FETs and I wouldn't recommend the 2030+ FETs unless you've got time to dial them in properly and the test equipment to verify their output is clean. You'll be quite frustrated with the experience otherwise.
Since your amp uses 8 FETs with 2 bias controls, you don't need all 8 to perfectly match. Two sets of matched quads is fine since each set of 4 has its own bias control. I'd start around 50mA per FET, so 200mA for each half of the amp, or 400mA in total. You can tweak a little either way to get the least distortion/best linearity without sacrificing power. I would imagine that the 8 FET amp would do 140-180W-PEP, based on my experience with the 4 FET amp. How much distortion you find acceptable is up to you though. I'm usually many db more picky than most hams. Don't take my long-winded post as gospel. I'm not a tech, nor an engineer, and my experience with RF circuitry has been purely from a self-taught, amateur perspective.
Also to note, heat output was very similar between the 520s and 2030+ at the same output power and input current was also very similar for the same output power. The 2030+ doesn't have some magic efficiency over the other non-RF power supply FETs.
Also, garbage in equals garbage out. If the radio is generating distortion, that distortion will be amplified the same amount as the desired signal. I use a Class-A "ultra-linear" driver stage between my bench signal generator and the amplifier under test to generate a very clean drive signal while doing my testing. I can generate up to 10 watts of drive signal from 1.8 to 30MHz and 5 watts up to 50MHz. The amp is then connected to the radio for final testing after it has been verified to work properly under "ideal" conditions. I know, overkill for most people, but I have OCD issues.
Good luck and 73s.