初步怀疑：下面两个參数引起：

SQL> conn / as sysdba
Connected.
SQL> show parameter lock
NAME

------------------------------------ ----------- ------------------------------
db_file_multiblock_read_count        integer     128
ddl_wait_for_locks                   boolean     FALSE
distributed_lock_timeout             integer     60
dml_locks                            integer     48420
gc_files_to_locks                    string
lock_name_space                      string
lock_sga                             boolean     TRUE
SQL> show parameter pre

NAME                                 TYPE        VALUE

os_authent_prefix                    string      ops$
pre_11g_enable_capture               boolean     FALSE
pre_page_sga                         boolean     TRUE
SQL> exit
Disconnected from Oracle Database 10g Enterprise Edition Release 10.2.0.5.0 - 64bit Production
With the Partitioning, Real Application Clusters, OLAP, Data Mining
and Real Application Testing options
[oracle@KXSALE02 ~]$ tail

分析过程：
1 提取awr报告发现等待事件
Event Waits Time(s) Avg Wait(ms) % Total Call Time Wait Class
CPU time   267   103.3 
os thread startup 48 22 462 8.6 Concurrency
control file sequential read 19,523 4 0 1.7 System I/O
log file parallel write 1,322  2 1.1 System I/O
control file parallel write 1,202 3 2 1.0 System I/O
2 Dump systemstat 10 发现问题
    ---------------------------------------------------
      [80 samples,                                         12:37:56 - 12:39:16]
        idle wait at each sample
      [1 sample,                                                      12:37:55]
        waited for 'os thread startup', seq_num: 50784
          p1: ''=0x0
          p2: ''=0x0
          p3: ''=0x0
          time_waited: 0.458091 sec (sample interval: 0 sec)
时间也比較温柔

3 在測试库上开启这两个參数发现：
close(8)                                = 0 <0.000008>
close(11)                               = 0 <0.000007>
read(10, "NTP0 13690\n", 64)            = 11 <0.681689>   
fcntl(10, F_SETFD, FD_CLOEXEC)          = 0 <0.000009>

4  在查看其他生产库，假设參数不一样。基本能够确定是这两个參数引起！