I recently had a client’s vendor run their name servers on a antiquated system which allowed for the zone file’s SOA TTL to be set at 48 hours. This means that the authoritative name server told all other name servers worldwide to check in with it to make sure that this zone file’s records were up to date every two days. This may sound reasonable, except that the SOA TTL was changed to 24 hours one day before the name servers were moved to another registrar which meant that not all name servers had been told to check in once a day and not every two days worldwide. As a result, the clearing of caches of worldwide for this particular domain was still blisteringly slow and we could not change over to a new website in a reasonable amount of time.
In an attempt o facilitate the transfer, I did something a little unorthodox and, perhaps, something that did not actually work, but I will outline it here. At the client’s new registrar, I created new custom name servers, pointing the name servers to a server in another country which had managed to switch over to the new authoritative name servers. Naturally, this caused an error in the dns, but it also yielded an error which quickly propagated. I would say that about 20% of servers worldwide switched over to the change in about an hour. However, propagation stopped at this point, the idea being that when they all switched over, I would quickly correct the error and, hopefully, the new authoritative name server change would then be registered over the old one. But as I said, this did not work. Retreating, I set the name servers back to the old registrar, waiting about ten days before switching over again before the new authoritative name servers propagated worldwide in about four days.