Buffering of DB output with *INLR=*OFF

I think this is a mixed RPG / i5/OS question, but I'm not entirely sure. Please send me to Midrange-L if I'm off topic.

One of our primary applications is misbehaving. Some fields on one of the main files are getting updated incorrectly, but I am unable to duplicate the problem. I have been able to determine that the problem records have *loval in a particular date field.

My solution is to write (my first) trigger program which will log the name of the entity updated and the program making the change. I don't have exact statistics, but I believe that many updates to this field occur during the course of normal operations. Since my *AFTER *UPDATE trigger will be firing many times, I want it to be fast, so I'm leaving my log file open and exiting with *INLR=*OFF.

I've noticed in my testing that the records I write to the log file in the testing job don't show up until the testing job ends. I'm using DBU to look at the file, in case that matters. I know that writes are buffered and that I can use FRCRATIO or FEOD, but I don't really want to override the buffered writing unless it's absolutely necessary.

What I'm looking for is recommendations on whether or not FRCRATIO, FEOD, or neither are the way to go in this situation. I don't know enough about write buffering to know whether it could cause performance problems in the given situation, say when a lot of people end their interactive jobs at shift change time.

I hope that most of that was relevant, and please let me know if I'm completely misunderstanding this.