I have been running Oracle 9iRAC databases for over a year, and OPS for several years before that. For 9iRAC in conjunction with TAF (Transparent Application Failover), we have had very good reliability apart from if one node panics and crashes where TAF does not work until the TCP keepalive time is reached (default 2 hours.... duh!). Aside from that one instance, I can report 100% reliability, including surviving database instance failure, listener failure, application server failure, etc. Will go to 10GRAC next year.
Why do you want to go to RAC? is the real question.
If you suffer from poor performing "crapplication" code, it will still be crap after you go to RAC, and you will have more nodes poorly performing than you had before! Also if your application code isnt set up for failover, then you may be asking for trouble.
There are two reasons to go RAC... high availability and scalability. For scalability, if you can upgrade your existing hardware, add more cpu/memory, etc, then do that first. If you cant upgrade it, then you can increase capacity by adding extra boxes in a RAC system.
I guess you would have a lot of testing to do! too!
My advice, be sure you really need to go to RAC and design everything up front for RAC. Dont try to bolt it on later. You would be better advised to do a box upgrade on your current server if you need more capacity.