strace may stop process

Some versions of strace may leave the traced process in a stopped state when you stop tracing. 
Check if your strace command suffers from this bug as follows:

$ sleep 100 &
[1] 9400
$ strace -p $!
Process 9400 attached - interrupt to quit
restart_syscall(<... resuming interrupted call ...> <unfinished ...>  <-- I pressed ^C here
strace: ptrace(PTRACE_CONT,1,133): Input/output error
Process 9400 detached
$ grep ^State /proc/${!}/status
State:  T (stopped)
$ kill -CONT $!
$ grep ^State /proc/${!}/status
State:  S (sleeping)

If you see the process you just ran strace on has a stopped state as shown above, immediately send 
a CONTinue signal to it. Otherwise unpredictable consequences could result. If you strace'd a 
critical Oracle background process, upon continuing the process, the Oracle instance is likely to 
crash.

To correct it, update the strace package:

$ rpm -qi strace	#note the version and release
Name        : strace                       Relocations: (not relocatable)
Version     : 4.5.16                            Vendor: Red Hat, Inc.
Release     : 1.el4_8.11                    Build Date: Tue 08 Jun 2010 03:19:10 AM CDT
...
$ sudo rpm -e strace
Password:
$ rpm -qi strace
$ sudo up2date install strace #use yum instead of up2date on RHEL 5
...
Fetching rpm headers...
########################################

Name                                    Version              Rel               Arch
----------------------------------------------------------------------------------------
strace                                  4.5.16              1.el4_8.11          x86_64


Testing package set / solving RPM inter-dependencies...
########################################
strace-4.5.16-1.el4_8.11.x8 ########################## Done.
...
$ sleep 100 &
[1] 23029
$ strace -p $!
Process 23029 attached - interrupt to quit
restart_syscall(<... resuming interrupted call ...> <unfinished ...>
Process 23029 detached
$ grep ^State /proc/${!}/status
State:  S (sleeping)


Ref: https://www.redhat.com/archives/redhat-list/2009-October/msg00190.html