Having a duplicate renamed HDU for RGL_PRIO must be valid in the input form, thus this HDU must be an item of the RGL_PRIO of the combo box, thus it must be present in the image library. Thus we have duplicates in the image library.

We can :

• copy and rename the HDU just before exporting it to the algorithm
• keep only one HDU, and correcting the file given by BSMEM to avoid non-unique HDU names. Avoiding non-unique names is done when you add images to the library, but it does not modify a result file which can thus contains non-unique hdu names.
• keep only one HDU, and correct BSMEM algorithm. Note that the presence of duplicate HDU names in the result file does not seem to impact OImaging, since it does not pollute the library.

That's too painful. I tested and BSMEM works with duplicated HDU (same HDU_NAME), let's wait.

Finally I prefer the 2nd solution: when calling the IR software, the input OIFits structure in memory is written to a temporary file (before sending it through http) so it seems the good place to fix the generated OIFITS file to be really valid (FITS + OIFITS + Fits image validation) and the writeOIFitsFile() can pre-process / adjust the generated nom.tam Fits HDU ...

See IRModel.prepareTempFile() to fix the oifitsFile given to OIFitsWriter.writeOIFits(tmpFile.getAbsolutePath(), oifitsFile);

It was not difficult to do it in selectHDUs since it was only testing if the HDU for RGL_PRIO was "the same" from the one for INIT_IMG.
Do you think sharing the HDU can cause problems in the rest of the software and that it is safer to do it at the very last moment (in prepareTempFile) ?
Also if we are to avoid duplicates when we send to BSMEM should we rework OIFITS file received from BSMEM to remove duplicates too ?

I am just proposing a possible solution to fix the OIFITS file (duplicate HDU + fix HDU_NAME only before writing file) that seems less complicated than fixing the all GUI to handle HDU duplicatton on the fly if INIT_IMG = RGL_PRIOR (same choice) !
To be discussed.

Closed and maybe revisited later after a review with software authors organized by @FerreolS