WAIT FOR A UNDO RECORD

Cause

In fast-start parallel rollback, the background process SMON acts as a coordinator and rolls back a set of transactions in parallel using multiple server processes.
Fast start parallel rollback is mainly useful when a system has transactions that run a long time before a commit, especially parallel Inserts, Updates, Deletes operations. When SMON discovers that the amount of recovery work is above a certain threshold, it automatically begins parallel rollback by dispersing the work among several parallel processes.

There are cases where parallel transaction recovery is not as fast as serial transaction recovery because the PQ slaves may interfere with each others work by contending for the same resource. With such a transaction rollback performance may be worse in parallel when compared to a serial rollback.  

Becasue of this contention and the perceived slowness and ‘hang’ like symptoms (the database may seem to hang, SMON and parallel query slaves may be seen to take all the available CPU), DBA intervention may be taken. If a large transaction is terminated, the the cleanup may also take time.

Solution

You can disable parallel rollback by setting the following parameter

fast_start_parallel_rollback = false

BEWARE: that setting this parameter dynamically can cause problems on a busy instance with a lot of active transaction work and it is safer to set this with an instance restart so as not to change the rollback strategy on active transactions.