I don't care about your 'personal sample'. A sample of four units is an abysmally small sampling. If you asked 4 people on the street if they think that ghosts are real, and two said yes, does that mean half of Americans think ghosts are real?
You quoted 'thousands' of units may have this issue. While you have no evidence to support this claim, let's use it for argument's sake.
Thousands out of millions is rare. That's 0.1%.
Contact Apple for your options. Contact a lawyer if that's what you want to do. No one here cares if you do that or not.
I have worked with hundreds of iPhone 3GS units in the past. None of them ever had this issue. However, I wouldn't say that the failure rate is 0.0%, even though that is my personal experience.
If I'm not going to use my experience with (and I just checked my records to verify) 200+ iPhone 3GS units to state that the issue is a non-occurance (and that's the rate of incidence in my sample: 0.0%), then you have no justification to use your 4 units to present it as a common occurrence.
That's how math and statistics works; larger sample increase the odds of accuracy. If we add yours to mine, then it's around 0.8%. Still rare, even with our combined samples.