I guess we've either deployed as new or patched up to around 20 customers or so to 6.4 with deployments across a broad range of models and different use cases.
6.4 is 8 months old now as a release, so i really wouldn't call it beta. In fact our experience with it is better, than 6.2. As it happens i've just been working through the night at a customer with 2000E and 501E clusters on 6.2.3 where both clusters lost several firewall rules during a lights out test, something i've never yet seen!
So whilst 6.2.6 is too new to say much about, our impression is that 6.4 at patch 3 is better than 6.2 at patch 5. To my knowledge, none of our customers have had to roll back from 6.4 and certainly our own internal 400D cluster has been fine on 6.4.2 and now 6.4.3.
The only thing that springs to mind with 6.4 is from upgrades and how this version deals with flow/proxy inspection compared to earlier versions which required changes to the configuration rather than it being any bug
When it comes to bugs it can often be a hit and miss affair. You could hit one in any release and i wouldn't say it's a given that 6.2.x is going to be more stable for your particular use case than 6.4.x so that's why i say you need to test the functionality that's relevant to you first before going into production.
Years ago i piloted a 600+ deployment of 60Ds with 5.2.0, and rolled out with 5.2.1 without a single hitch. 6 years down the line and having gone through 5.2, 5.4 and now 6.0 i still haven't hit a confirmed bug and have only had 2 boxes die on me, which i would say is a pretty good record. Yet in other cases i've seen no end of problems with a single unit or cluster.