how about adding some delay method or postpone or similar idea to the mirror structure. When mirrorbrain adds a new file it should not answer yet requests for that file and only say n minutes after initial population. Wasnt mirrorbrain actually capable of making sure which file had which live and available mirrors, so why serve files at all that have no mirrors at all. I understand that the main source would be the suse server and as a last resort, but the whole project or company could not live with that situation I guess, so lets assume that having no mirrors for a file is a non-acceptable state, so just do not publish that information out to asking clients. Only serve files that have mirrors available. This a minute delay or however mirroring infrastructure out there works, when and how those mirror nodes fetch the files themselves I am not familiar with. Regards.