Avg user wait time for task <task_ID>: <duration> ms, <number_of> waits, application pid <pid>
In a lock collision, the wait for a lock to be released is very long.
This can be caused by a transaction that takes too long. Long wait times (task state vwait) can also occur when many applications want to lock the same object.
If the locks are requested by other sessions, the database system attempts to execute locking tasks with higher priority within the database kernel.
For information on the current locks, see the system tables SYSINFO.TRANSACTIONS and DOMAIN.LOCKS of the database instance.
Check the following aspects of the application:
Is it suitable for isolation level 0 (uncommitted)?
Can the time between the lock request and the COMMIT statement be reduced?
Can you avoid simultaneous locks on the same row by splitting tables?