Freeze and cut often for release?
I've read from people wanting to generate CUTs from unstable, if testing is frozen. However, maybe increasing the frequency of CUTs during the freeze instead of testing (weekly, or manually triggered even faster) may actually contribute to shorten the freeze time for testing before a stable release. Presumably, this way a significant user base following the "known to work mostly" CUTs will help in a thorough "last weeks" testing of the next stable release (burst update for stable).
I'm thinking of the CUTs as versions ("snapshots") that the devs consider worthwhile for users to check out and use, if they consent, want, or are in need of always having recent versions installed. The users trust the devs to decide and are willing to live, not with the cutting edge itself (unstable/current), not with what is aggregated by the dev community (testing), but with the fresh cut, harvested by the devs.
If testing is frozen and itself considered as "known to work mostly", the CUT could even be made a direct symlink to testing during the freeze; making all fixes hit the CUT users directly.
This should always allow people to use and test reasonably recent versions. The knowledge and experience of the devs provide some level of shield against larger breakages, while the CUT users help debian in testing.
For the cutting edge itself, there is unstable and its related (currently derivative aptosid) user community.