Certain administrative and operational changes — changing registrars, changing DNS servers and, if outsourced, DNS operators — have always had the potential to cause temporary name resolution failures. With some planning, usually involving lowering TTLs on some or all records in a zone in advance of the change, it has been possible to minimize if not obviate such failures.
DNSSEC adds complexity in that signatures also have lifetimes and some administrative and operational changes require re-keying. If not done correctly, such changes can cause signed zones to fail to validate — to go dark — for longer than desired or expected.
Internally, within the DNSSEC Deployment Coordination Initiative, we’ve described the goal as being a ripple-free transfer, and have made presentations on the topic (e.g., SATIN 2011 180KB PDF). Done properly, there is continuous, secure resolution throughout the process — no need to have a zone go unsigned/insecure or fail to validate at any point.