I've spent a bunch of time researching this. I use a pair of EPP-400-36 in an audio amplifier, and other than the low-load whine they work pretty great.
If we take look at the
RPS-500-c datasheet (same as the EPP-500 but medical approved), we see an airflow diagram provided. 25CFM directed at side, looks to be a 70mm to 80mm fan 15mm thick, in a push configuration. The top fan version is offset, and does not mention a CFM, but it show a pull configuration. The enclosed with 40mm fan has it pulling through the primary side and offset 10mm from the PCB. Noctua 40mm have an airflow of 5.3CFM, 60mm have an airflow of 17CFM, 80mm have an airflow of 31CFM. Noctua 80mm can do it, but would need to be tested. The
Meanwell RPS-500 test report lists burn-in temperatures at 320 watts/45c ambient, and It looks like your main area of concern will be the transformer (where all the airflow is directed in the diagrams) plus several of the switching mosfets. It does not look like the backplate cools the underside of the PCB with thermal pads to components but I don't have one on hand to confirm. You might gain from adding heatsinks onto the existing heatsinks, but they
must be isolated with thermal pads. These do have OTP, so you should be safe doing airflow tests in a hot case.
A top mount 15mm thick fan on the RPS-500c will bring you to 58mm. On the RPP/EPP 500 it's 56mm, and on the EPP-400 it's 50mm plus stand-offs. going with side mounted 15mm gets you to 90mm wide, plus adding height if using greater than 40mm fans. Either way is still less than an SFX PSU, but it starts to get close.
Is this for 12V only systems? If using one of the 12v DC-ATX adapters, how do you plan on getting around the power limits on those? I was thinking of grabbing a picopsu and then feeding the the GPU/CPU power connectors through relays timed from the motherboard PS-ON signal. This leaves the meanwell always-on. My other idea was to build out the logic to make the mean well PS-ON compatible with motherboards and add a pair of SIP DC-DC converter modules (20$ish each for 15-20A from mouser). I'd need to doublecheck the ATX timing spec to make sure this will work.
There are a couple parts about this PSU that take it out of ATX spec. Its ripple is 200mv, test report shows 187mv at full load. PS-ON is inverted, and PS-good may trip early (though under full load the report shows 118ms, which is within spec). hold up time is short at 220v but fine with 115v. It also need to be run class 1 to hit consumer EMI class B spec (grounded through standoff's to a metal enclosure).
Personally, I think the big advantage of these guys is the 1RU height. In my mind the perfect way to set this up is by making sure there is case airflow, add some small heatsinks to the heatsinks, and building out the support circuitry to the side, making the overall size a little bigger than a 120mm fan (double thickness). My guess is with a little case airflow most systems that pull 300w on load should be just fine. If need more, you should be able to tie two of them together in parallel and always on and use the pico + relay setup for CPU/GPU cables. That would let you run passive and give you some headroom for running in a hot case. Size would end up 127 x 151 x 41 plus relays.