Description:
When importing the attached documents, try to read them with VoiceOver on MacOS when they appear.

Results:
1. The .docs file reads the first line, but doesn't read many of the others.
2. The same named .odt file is the converted .docs file, with some additions. As soon as down arrowing into these additions, reading becomes better, but VoiceOver reads each paragraph as a new text field accessible, e.g. treats each new paragraph as if it was a true focus change.
3. The third file, the .odt file, is an ODF export from EtherPad. It also has some additions at the bottom. Notice how VoiceOver reads the first bit, which was the original .ODF exported, as a single unit, treating new paragraphs no differently. But as soon as the additions are reached, again treats new paragraphs as separate accessible entities.

Steps to Reproduce:
1. Open the attached documents in turn.
2. Try to read them line by line while VoiceOver is turned on.
3.Notice the difference when moving into the lowest paragraphs.

Actual Results:
VoiceOver reads many lines incosistently or not at all. Treats new paragraphs as separate accessible entities, including pretending that focus changes happen.

Expected Results:
It should read all lines, treat all paragraphs as just lines separated by blank lines in one accessible entity. Or if it needs to separate paragraphs into their own accessibles, do so consistently for _all_ paragraphs.


Reproducible: Always


User Profile Reset: Yes

Additional Info:
MacOS 15.2 sequoia, but I remember seeing this behavior much earlier in earlier LO versions, but just now methodically reported this since I want to start using it more. And it really hinders productivity and sometimes makes it impossible to work with documents at all.