GATHER_STATS_JOB behavior changed between 10g and 11g

We have this system where materialized views are scheduled to rebuild within the same maintenance window in which Oracle runs the automatic GATHER_STATS_JOB.
Not the best idea, but in 10.2.0.2 it did not cause any issues.

After upgrade to 11g, GATHER_STATS_JOB causes "ORA-00054: resource busy and acquire with NOWAIT specified or timeout expired" messages in our alert log.
It runs much faster though :)

I can completely understand the new behavior - why waste time (and possibly hang the stats gathering job) when you can just skip resources. However, we'd rather have the old behavior back.

Anyone knows how to get it back?

Thanks,
Chen Shapira

--
http://www.freelists.org/webpage/oracle-l