The short answer to the how much load question is 'enough to get everything hot.' A wally-world Chinese gasoline generator probably shouldn't be run sustained at rated output very long. I'd probably do the test run (30 min or so) at 2/3 rated load and do a 30-60 second full-load check just to make sure everything could cope, then back down to 20% load for a few minutes' cool down before shut off.
By comparison, the mil-spec generators are extremely under-rated. As sewerzuk's test case showed, the MEP-005A in good health was capable of pulling 166% of rated output. I haven't loaded my MEP-002A that hard, but I did have it reading 100% on the load meter with a pure-resistive load, which should be about 120% of rated load. I didn't have any more space heaters to load it with, so I don't know how much more was in the jar.
That said, wet stacking shouldn't be a problem even at 50% of rated load as long as the test run is long enough. My generator doesn't really start to get warm until it has run under load for half an hour or so. I'd say that the air-cooled diesels probably need to run an hour at a shot to get everything up to thermal equilibrium. I have not observed that the warm up happens a lot faster at 100% rated load than it does at 50% of rated, which suggests that the thermostat on the cooling tin is doing its job.
Probably the worst cases of wet-stacking I've ever heard of were on the British Rail Class 55 Deltic locomtives. There were two
Napier Deltic engines in each one, and Brit Rail had trouble finding heavy enough trains to really work the engines. Combine that with a complicated and oily two-stroke diesel exhaust system and lots of engine idling, and you have a recipe for the famous '
Deltic exhaust collector drum fire' that resulted in quite a few of them going to the shop for repairs during their operational lives.