This problem can cause a server stop or SQL errors like:

2937:20110721:120009.940 [Z3005] query failed: [1062] Duplicate entry '22173-1311246000' for key 'PRIMARY' [insert into trends (itemid,clock,num,value_min,value_avg,value_max) values (22173,1311246000,10,0.000000,0.034000,0
.140000),(22176,1311246000,10,0.000000,0.000000,0.000000);
]
2937:20110721:120009.941 [Z3005] query failed: [1062] Duplicate entry '22172-1311246000' for key 'PRIMARY' [insert into trends_uint (itemid,clock,num,value_min,value_avg,value_max) values (22171,1311246000,10,1549,1812,2255
),(22172,1311246000,10,873,4341,15906),(22174,1311246000,10,541,1184,1760);
]

I can't reproduce it, but the problem is quite real!

----------

Valgrind output (trunk r25576):

==00:00:04:56.530 23549== Conditional jump or move depends on uninitialised value(s)
==00:00:04:56.530 23549== at 0x42F78A: DCflush_trends (dbcache.c:473)
==00:00:04:56.530 23549== by 0x431B47: DCmass_update_trends (dbcache.c:712)
==00:00:04:56.530 23549== by 0x43301D: DCsync_history (dbcache.c:2105)
==00:00:04:56.530 23549== by 0x4121B9: main_dbsyncer_loop (dbsyncer.c:65)
==00:00:04:56.530 23549== by 0x4115CA: MAIN_ZABBIX_ENTRY (server.c:710)
==00:00:04:56.530 23549== by 0x43F120: daemon_start (daemon.c:286)
==00:00:04:56.530 23549== by 0x411B69: main (server.c:469)