This a simple server with two domains and one email account per domain. Mail is set up to use SpamAssassin and Greymail on both accounts.
Autolearn on both accounts appears to be operational from the headers and the accounts have been operational long enough to have accumulated enough ham and spam results for Bayes to be able to do its stuff.
One account shows Bayesian filter working OK in the incoming mail headers, while the other does not (and possibly never has). Settings appear to be identical for both accounts, so what have I missed? Suggestions would be appreciated.
Environment: current GPL VirtualMin on Ubuntu Server 12.04