At my 3rd world team we used large (at the back) and small (front) diameter DGBBs for years. I won't say we were a top team but at least we were consistently in the top 15 in AutoX. I never saw a bearing fail from 2007 to 2010 (and plenty of kms in between). I have even jumped on the car (have pictures, bumpy track, landing sideways sometimes BTW, 3 to 4 inches for about 15 laps over and over again). I have hit countless types of bumps. In MIS 2008 the endurance track forced you to hit the oval bank diagonally, really hard (right front felt like hitting a curb or worse). wing teams were even advised that they would rip the wings off because the angle was significant. On the car it felt like a hard impact curb (but for longer time since the bank doesn't level as a curb does). In MIS 2010 again, but not so bad.
I won't say exact code but they were about 130g each, inside diameter in the range of what people have put here. So with that evidence I can confidently say that you can go light, strong, and stiff. I won't say they were cheap (they were NOT $20 each), but (as with the rest of the car) you have to choose between weight, strength, and reliability. It is not like they costed $20k so in the end I would think cost should not be a problem for bearings, as long as you don't go crazy and buy some super bearing. To save cost I would say go to sheet metal welded uprights. You can get them light as hell, and cheaper than your cereal.
My guess and advice would be that, if you choose DGBBs (regardless the weight and maybe even the load rating, for the sizes we are talking here a large group of options would handle the loads), the real reliability comes from the design concept: how they go in, tolerances, preload, axial spacing between them, locking system, how stiff their housing is, etc. I have seen really bad designs in some cars and in such case the problem is not with numbers but with concepts, and there is nothing a super bearing will do for your compliances.
Hope I helped in some way.
JP