At first when I opened your ODF sample, I got everything right, without the spurious white space at bottom of page. My computer is Fedora 43 (Linux) with LO 25.8.3.2 but your fonts are not installed and are therefore substituted.
It was difficult to compare with the PDF because the ODF sample is not the same as the PDF. But I noticed that the size of the cells was not identical: Hebrew text was shorter here (effect of font substitution?) and was always smaller than English text. So, I began to fiddle the table. When I enlarged the English column, I got your no-row-split behaviour but I could not fix it by reducing column width. It looks like once you have the no-row-split layout, you can never revert back. This is probably a bug.
I looked more closely to your table and text formatting.
You table has been tweaked manually. Its alignment is no longer Automatic but Left with a 0.02cm unused space. Text direction has been set to Right-to-left (RTL). This has the consequence of reversing column numbering which confused me a bit (but not for long).
I changed table alignment to Automatic and text direction to LTR (which is consistent if your main language is en_US as reported by the bottom status bar). Layout was still OK. I forced narrower Hebrew column width so that its text was taller than English. I entered “no-row-split”. Since I already experienced weird behaviour in tables (within multi-column sections), I saved and reopened. On opening, layout was as expected.
I then remembered a trick which solved my issues with erratic cell breaks: press Enter followed by Bksp to eliminate it. Usually the table is reflown to its expected layout.
In your sample, this didn’t work. But something weird happened: text flushed from below reappeared on the previous page.
So I better looked at the screen. The note separator is on the right side. I went to Default Page Style to discover it was set for RTL, suggesting your document is mainly Hebrew. I number the pages to have an idea about page flow. Obviously, your document is RTL and it remains RTL no matter what I try.
Multi-lingual document is a difficult art. Here you mix RTL and LTR in parallel columns. The “separate” flows have contradictory and conflicting requirements regarding page overflow (parity). I think this is the main cause of the problem.
IMHO, the only way to solve it presently is to tune your column widths for the best trade-off, minimising the number of splits.
I also observed a lot of direct formatting (DF), which does not help. As I said, multi-document are difficult. You should design ad hoc styles for your fragments, instead of applying DF over them.
I noticed that some rows did not contain the same number of paragraphs in the cells. Usually, translations are paralleled on paragraphs which are the basic units of rhetoric significance. Of course, provided the notion of paragraph has the same purpose and relevance in both languages.
