can be closed for any use even for days, if required, and make the changes.
Actually, no, they can't. You have to schedule upgrades literally weeks in advance and get permission from people at Admiral rank and not always just a two-star. But we've beaten that horse so I won't flog it any more. You know it is bad. You know there is a risk. So let's try to figure out your problem.
In your dialog with Arnel, he suggested using a method that would directly build a link from the FE to the BE. You had mentioned a routine that you used to make structural changes in the BE. What we need to do is figure out when this difference slips in. You say you have done a sequence of Delete FE Table link; Compact & Repair (C&R); Create FE table link - and that sequence causes a difference. So specifically when you do this, how do you create the FE link after the C & R operation? Your code, a ribbon manual operation, or code like Arnel suggests? OR was your code more or the less the same as what Arnel suggested?
The reason I'm asking the question this way is that deletion of a link followed by C&R should leave no vestigial traces of any previous incarnation of the deleted link. That is because of the way a C&R works. What you have after that step isn't even the same file. It is a structural copy of the currently active content of its predecessor file, omitting all obsolete, deleted, and inactive content. But it's not the same file because during the C&R, both the original and the new files coexist.
Therefore, if this happens, it cannot be due to any content in the new version of the file; it has to originate with something inside the old file that stays visible and active even after the link removal.
I also want to avoid making an assumption. When you say that you have tried to delete the link, C&R, and rebuild the link, that obviously is on the FE file. But have you separately tried to do a C&R on the BE file? Because the link HAS to use data from the BE file, so I'm wondering if looking at the FE file is looking in the wrong place.