We have a homegrown tool for rough cut editing which is accurate only to about 1/2 second; meaning we have to tweak each edit point. About 20-30% of our day is spent tweaking those edits. If we can get frame-accuracy, that work goes away - allowing us to work more on finishing. Assuming we use the same frame rate as the original media, will the rough cuts be frame accurate? Does that carry over into CatDV-assembled timelines?
A lot depends on your proxy format of course. Most web proxy formats aren't frame accurate themselves so you are limited by that. I've not pushed it but it's certainly not a half second, more like within a frame or so and from what I've seen on ProRes, it's right on.
CatDV is a surprisingly good little editor although almost none of my clients use it for that strangely. (They tend to be more focused on logging and traditional MAM functions.)
We use CatDV's editing function all the time, with a writer roughing out a sequence which gets moved into FCP and opened fully populated with proxies. I don't notice any accuracy issues, but then I'm not looking for any. I'm recutting and adjusting anyway. I'm a lot more concerned that when I'm finished with the proxies and I load up the masters I'm going to have the same cut. And I do. I use motion jpegs for proxies and work frame to frame. They're tiny (and ugly) and I get to store ALL our footage "online" all the time so I always have access to everything. And I get a one-to-one relationship frame to frame between the proxies and masters. This is really our greatest use and value of CatDV.