In the past, if you wanted ORA errors to be properly reported you needed to leave the Array Size at 1. That way if it said you had three duplicate key violations and dumped the rows back into the log, you knew there actually were three and those were the ones. Bumping the array size up would cause 'mismatches' with reality. One error might be reported as three and the rows bounced back down the reject link may not be the actual problem children.
In the past, I haven't worried alot about this, we routinely do 'all or nothing' loads and the first error aborts the job. Now, a new project wants to allow a certain level (up to a threshold) of these errors to be captured and shoved into an error reporting system. Leaving the array size at 1 for some of the loads we're doing makes things rather... slow. However, I'm afraid that bumping it up for speedier loads may make the error reporting 'less than reliable'.
![Confused :?](./images/smilies/icon_confused.gif)
I was wondering if anyone has had similar experiences or how they are handling things of this nature.