When new mail is checked on IMAP accounts (both SSL and non-SSL) and the remote host is unreachable, sylpheed-claws freezes on the read() syscall until the host becomes availabe again. That makes sylpheed-claws totaly unusable (except for hand-editing config files) when new mail is checked on start-up and new-mail is automatically checked for the unreachable host. This is not good, it should timeout. And maybe ignore that host until 'Get' is explicitly requested for it. I have not checked how POP accounts are affected. Remote systems with IMAP servers are Linux and FBSD 4.8. I have no info on what IMAP software runs there.
Try setting this option to an acceptable value: "Configuration->Common Preferences->Advanced->Socket I/O timeout"
Apologies. What I meant was: "Configuration->Common Preferences->Other->Socket I/O timeout"
1.9.11cvs32 should make that much better.