we're a small school. only 1 zimbra, 1 clustered active/passive gw po. 400 users and resources total over next week and a half, 150 users this weekend... including 1 president, 3+ vice presidents, deans, etc. with complex interlocking calendar shares.
rough start. with help of zimbrans in india (highly recommended), determined my strategy of parallelism was wrong. ended up causing lots of errors due to mysql and attachment text extractor/indexer contention. actual best strategy is careful watching of only 2 parallel tasks - each runs faster, giving optimal overall throughput.
daylight savings bug on repeating appointments was *not* fixed in time for us. we hope, but are not certain, that we will be able to zmmailbox ef /Calendar and re-import just calendars. else, we will have to ask dept assistants to correct our VIPs' calendars manually.