hmm.. are you HUP-ping the correct pid? you could possibly truss syslog-ng while sending the HUP to see whether it closes and reopens its files. (it should do so)

The truss showed only /var/adm/messages was being reopened. I found that there were multiple syslog-ng's running. Killing them solved the problem. I falsely assumed the solaris init.d start script checked that. Thanks, -- Wim Livens.