F9860 is apparently the driver:
Apr 14 15:22:23.724049 - 5588/1012 WRK:Starting jdeCallObject SELECT * FROM OL920.F9860
Apr 14 15:22:23.724050 - 5588/1012 WRK:Starting jdeCallObject ORACLE dballInitRequest OCIStmtPrepare for 329a0428 stmthp
Apr 14 15:22:23.727000 - 5588/1012 WRK:Starting jdeCallObject Exiting JDB_SelectAll with Success
Apr 14 15:22:23.729000 - 5588/1012 WRK:Starting jdeCallObject [PKGBLD] - Wed Apr 14 15:22:23 - Start the insert of objects into the Repository Tables: F98780R and F98780H.
So it's not immediately obvious what exactly in F9860 it checks to decide whether to include an object or not.
It then checks status in F9861 and finally does this, per object:
Apr 14 16:00:52.274009 - 5588/1012 WRK:Starting jdeCallObject SELECT * FROM OL920.F9861 WHERE ( SIOBNM = 'B0000011' AND SIPATHCD = 'DV920' AND SISTCE = '1' ) ORDER BY SIOBNM ASC,SIMKEY ASC,SIPATHCD ASC
And the last one, for instance, it does for B0000011 & B0000013, but not for B0000012, so it's missing from the final F98760* tables as well it's sources are not in the package folders.
There's not a single mention of B0000012 in the entire ~3GB log.
I'm intrigued by SIOLCD04 here. But I do not see what's different between B0000011, B0000013 and B0000012 looking at F9860 alone ;-( Apart from SIOLCD01 value, but how can this matter??? - I have the same data in another E920/TR924x and it does not cause such issues.
Baffling...