We have a NETWORK discovery and an operation action for this discovery which adds hosts to a group and link an template.
The template contains one LLD discovery rule with one item prototype.

We consider that discovery rule find 4 hosts.
When server links the template it inserts extra and wrong records to the "item_discovery" table.

extra&wrong (1st in the inserted data) - mapping from the host item prototype to the template discovery rule. Frontend do not do that.
previous extra&wrong + extra&wrong ... for Nth inserted data (count of inserts increased for every newly discovered host) - mapping from this host's item prototype to the previous, already added host's discovery ruleS

Sorry, it hard to explain

How it looks for 4 just discovered hosts:
14341:20120607:205729.391 query [txnlev:1] [insert into item_discovery (itemdiscoveryid,itemid,parent_itemid) values (1061,26575,26572),(1062,26575,26574);
14341:20120607:205729.426 query [txnlev:1] [insert into item_discovery (itemdiscoveryid,itemid,parent_itemid) values (1063,26577,26572),(1064,26577,26574),(1065,26577,26576);
14341:20120607:205729.481 query [txnlev:1] [insert into item_discovery (itemdiscoveryid,itemid,parent_itemid) values (1066,26579,26572),(1067,26579,26574),(1068,26579,26576),(1069,26579,26578);
14341:20120607:205741.525 query [txnlev:1] [insert into item_discovery (itemdiscoveryid,itemid,parent_itemid) values (1070,26581,26572),(1071,26581,26574),(1072,26581,26576),(1073,26581,26578),(1074,26581,26580);

Count of rigger prototypes also shown as increased the same as for item prototypes.

The discovery rules which have "+ extra&wrong" mapping are in error state (visible in the GUI):
"Cannot create item [vfs.fs.size[C:\,pfree]]: item already exists Cannot create item [vfs.fs.size[D:\,pfree]]: item already exist"

Filtered (only of discoverer process) debuglog attached. The queries from this debuglog.
Try to imagine scanning of /16 subnets - some hosts have thousands of item and trigger prototypes.
System becomes unusable, database is broken!