Here's my situation, using SSIS 2005.

In a Data Flow task, I have an OLE DB command component that calls a
stored procedure which in turn updates some values in the target
database. If that stored proc fails and causes an error, I need to log
that error and continue processing the rest of the rows. So, I set up a
failure precedent constraint from the component to a script task
(transformation) which generates an error message and then passes it on
to another OLE DB command component that logs the error in the
database. When establishing the failure precedent, I made sure it was
set to redirect the rows in the case of an error.

To test this out, I altered the stored proc to raise an error every
time it's fired so that I could make sure the error was being logged.
The problem is that my rows aren't being redirected down the failure
path set up by my precedent out of the OLE DB command component.
Instead, processing just stops when the errors are encountered. I can
see in the Execution Results tab that the errors I raised in the stored
proc are bubbling up. I can also see the following from the
DTS.Pipeline "[DTS.Pipeline] Error: The ProcessInput method on
component "Update Rows" (244) failed with error code 0xC0202009. The
identified component returned an error from the ProcessInput method.
The error is specific to the component, but the error is fatal and will
cause the Data Flow task to stop running. "

I've set the MaximumErrorCount for the Data Flow task to 9999 to no
effect. My question is, why aren't the error rows being directed down
the failure path when the error in the stored proc is raised? Why does
processing just stop and the Data Flow task fails? What am I missing?