If it is my suggestions that you are thinking of then I think it would be to base the number of FREELISTS (not INITRANS) on the maximum observed interested transaction count. The rationale is that in the absence of multiple freelists, all concurrent inserts go into the same block (normally at the high-water mark) and the number of ITL entries dynamically created in each block reflects the number of concurrent inserts. Therefore, that is the minimum number of freelists that would be needed to avoid freelist sharing and the attendant risk of buffer busy waits. My opinion on INITRANS is that you should almost always use the default value.

You are right that finding the maximum ITL count technically requires you to dump the entire table. However, given that this is intended to address the issue of concurrent inserts at the high-water mark, I am normally content to dump say 100 blocks just below the HWM, and work with the assumption that what I see there is representative of the rest of the table.

I believe I've seen suggestions to base the value for INITTRANS on the maximum observed ITL for the object. Doesn't this require dumping every block for the object to examine the ITL? Any short cuts?