A 30MB mailbox, while large, is not ridiculously large. If the mailbox is in traditional UNIX mailbox format, it will have to read the entire file; but that should take no more than a few seconds unless I/O is very slow on your system. Using mbx format will speed this up greatly.
100 seconds is absurd. It shouldn't take that long to locate the messages in a 30MB mailbox. It might, however, take that long if the IMAP client does a full RFC2822/MIME parse of every message in the mailbox. Have you tried a client, such as Pine, which is known to use a single persistent IMAP session, and does not fetch message RFC2822/MIME data until (and unless) it is needed? I don't know what Evolution does. Many webmail based clients will spawn a new IMAP session per click, and it is possible that repeated reparsing of a 30MB mailbox exceeds your system's load threshhold. One thing to try is to trace the IMAP protocol negotiations and see what the client is doing. Often, in these sorts of situations, you will find the client doing something silly. If you are running UW imapd, and it really takes 100 seconds for it just to do "SELECT INBOX", I certainly want to know about it. I'll bet though that something else is happening.
