I'm preferring the 2G splits because they work without issue. I've had some issues with 4G splits during extended testing - although I'm not running the very latest rev of firmware.
Be sure to set your preroll buffer to the highest setting.
What were the issues 4gb split, other than not being able to open in programs other than CD Wave? Personally, I like the 4gb split b/c I don't have to merge anything before tracking in CD Wave. I import directly into CD Wave and move on, bypassing the merge step as much as possible (with the exception being the 2+ hour set)
I ran continuous 24/96 recording tests for about 17 days, each test being a freshly formatted drive set to record until the disk was very full (between 29 and 37 hours depending on the test) and then start over. During the test I would stop and restart recording, cycle power and restart recording, navigate the HD filesystem and try to play files, etc. All tests were run with external power so that a battery related shutdown could not be blamed for a FS error (although that is another test set entirely).
It had been widely reported that there was a bug that caused the 722 to lock up when attempting to play 4G files. What I found was that I could not get though an entire 37 hour session (which was really like 42 hours considering the various interuptions) without the 722 locking up when I recorded using 4G file splits. The system would generally do fine, regularly lock-up when I tried to play a 4G file (not always), but it would occasionally lock up during recording. In these cases IIRC the counter was incrementing but there were interrupted files and the system would often require removal of all power to recover. In some cases, when I hit <stop> and then tried to restart the sessoin, the subsequent <rec> would lock up the 722. I had only one trial that did not exhibit any failure over 37 hours using 4G splits. Once a particular 4G trial had any sort of failure, subsequent recordings were more likely to fail and the system seemed mildly unstable. The "slow" message being reported regulary during the failures.
When I ran the same tests using 2G splits, I never had a single problem - completely flawless performance. So now I only run 2G splits, even though it isn't what is best for me.
I reviewed the posts here to see if others were having the same problem but most people seemed to be starting each session with freshly formatted media. In my tests, I had one case were a freshly formatted drive had an error early enough to have been during a long show, but most of the failures began to appear about the 8h mark and some not until nearing 30h mark. So I can image that someone running a reformatted drive for each sessoin may never suffer a failure.
I originally suspected that it might be heat related but ruled that out by reproducing the failures at ambient room temperature, then running the heat test twice using 2G splits with no errors even thought the 722 got hot. So it was not related to heat affecting the HD.