Created attachment 128143 [details]
Obfuscated .odt files produced from the MS files, that reproduce what I have described above

I have a novel; I decided to make sure every mention of the thing called "it" in italics, should be called "It" in italics (upper-case "I"). I have 3 versions of the MS: one for the 4"x7" print edition, one for the 5"x8" edition, and one for the Kindle edition.

I opened all three files.

I performed the Fine & Replace on the 4x7 file with no problem: I selected Whole Words only, typed in "it" (no quotes) in the search field, "It" in the replacement field, clicked on Format and left the font unselected but chose style Italic. I carefully made about 15 replacements (about 4 possible replacements would have been wrong in the context). F&R worked as expected, finding only "it" in italic as a whole word.

I then moved to the 5x8 edition MS. In the dialogue, most F&R fields were already set from editing the 4x7 file; but I had to type "It" into the replacement field.  I then started searching.  But now, F&R did not work correctly: it would only find regular-format "it" words.  I watched on one page, where "it" was in italic, and another "it" appeared a little below.  It would only find the 2nd, not the first.  but after that, it would only find regular-format "it", and refused to find italic-format "it". I notice that under the Search field, the words "Italic, normal," appear, for the file where F&R is finding the wrong-format text.  In the file where it is working, the text is displayed without a trailing comma: "Italic, normal".

In the Kindle edition (see file WT-CS-KDP.odt), it found a few italic "it"s, then started finding only regular "it"s.  F&R seems completely unreliable: I just tried it, and it found in succession about a dozen italic "it"s, bypassing regular-"it"s as it should, then started finding regular-"it"s and ignoring italic "it"s. Basically at random.  The function simply isn't working reliably.

I cannot supply you with the original source documents, however, as Amazon's T&C would not permit me to do that. So I have obfuscated the files, but leaving the letters "i" and "t" and untouched, and saved those files and attached them; they do evidence the bug, for me.

This bug may relate to 95633, 99038, 31491, or 95927, but none of them really sound the same as this, to me.

LO is becoming increasingly frustrating to me, as an author.  This is the 3rd bug which is preventing me from making a specific edit to my LO MS to correct an edition.  In this case, at least I have a workaround: I can Undo each change in the file for which F&R worked, carefully note where that was, find the same place in the other editions, and manually make the correction.  For the others (103078 and 62603), they're show stoppers.  It's very disheartening.

[Later...]
Having just done the manual fix-ups, I'll also note that Undoing to find them all worked well (there were actually 13 occurrences). But that there is also a bug in Redo, as after each redo, the text was converted from italic to regular.  Fortunately, I saved a copy of the file before I started Undoing, so this is not a major calamity.