Thanks one more in advance for any help/critique/advice.
I didn't state clearly my intentions, so to correct:
The goal of the test is to check durability (lifetime). To be more precise - with a 5000 km maintenance period, which grease fares better and for how much (wear test by weighing and visual inspection after a thorough cleaning).
Rolling resistance is very small for any properly lubricated bearing on a bicycle. Oil would beat practically any grease on those tests, but fail at service intervals. So I won't even be measuring that.
As for bearings - I'm thinking 6903, of decent quality, since the machine mount is 16 mm diameter - it would make it easier to produce an adapter. Would that work fine - it is used in non-Shimano wheels for all I know? Get a batch of 10 for a start, from a good quality manufacturer, clean them all, weigh them all, then put them in an oil bath to prevent corrosion since I can test only one, perhaps two at a time at most.
The test should give results that can be used for cup and cone bearings as well IMO, unless I've missed something.
For shock tests, all I can use for now is a Charpy pendulum, which requires removing the tested bearing off the test machine, hitting it
then putting it back. I've already got a DIY fan if it is required - 4 PC cooling fans, fixed to a single rack.
Not sure how to make a DIY hitting gizmo. I'll post pics after the holidays (mid August). Then I'll know more about lab time and all the resources available. The best possible option would be for someone who actually knows what they're doing to take it up as a thesis, or a grad project. But for now - they're happy to let me use the lab for free, with me providing all the samples and doing the testing, which is more than awesome. For now I want to do all that is up to me to make it be realistic and to prepare all that's needed.