We have set up Oracle CDC against several Oracle instances with the same database structure. Occasionally, we see the error referenced in the subject of this post. When it happens, change capture is sometimes suspended completely (status shows LOGGER), and sometimes continues without an apparent problem (status shows PROCESSING), however even in the latter case, the "failed to position" error appears in the trace and in the Application log every few seconds. From looking at the diagnostics info, it seems that only one thread is processing changes, while the other thread is "broken" and raises the error every few seconds.

Oracle version:
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

The diagnostic info generated from the MMC console is attached.OracleCDCDiag.zip

Please suggest how we can troubleshoot this further.