A few kind people reported to me that SQLd360 disconnected in the middle of the execution. Looking into the details the cause was an ORA-7445 evaopn3() but such errors are a little tricky to track down because they only show up with a combination of faulty plan (that’s the bug) with specific data, that’s why the error never reproduced in my labs as well as at several other folks’ systems.
Galo Balda was kind and patience enough to report the error and test my suspects (Thanks Galo!!!) so we were able to identify a way to avoid the error.
The ORA-7445 was caused by bug 12672969 (I suggest you check the details in MOS, the bug is pretty “common”) and starting today SQLd360 works it around so that your execution shouldn’t disconnect anymore!
If you tried SQLd360 in the past but ignored it because of the unexpected disconnect then I suggest you give it one more shot now that the error should be taken care!