[cybmole]

I converted a retail azw3 to epub - with my usual preference settings ( heuristcs OFF)

but the conversion took a very long time, the resulting epub took even longer to open in sigil , and it seemed that the reason was that some chapters had been split into many, many XHTML files, with typically only 1 sentence per file.

I re-ran the conversion with all structure detection settings blanked out & it was fine. i.e. I removed this

//*[((name()='h1' or name()='h2') and re:test(., 'chapter|book|section|part|prologue|epilogue\s+', 'i')) or @class = 'chapter']

and removed this

//*[name()='h1' or name()='h2']

& with those bits removed I ended up with 1 file per chapter, & "normal" conversion /load times.

I don't do a lot of azw3 to epub, so I don't know if this was a specific issue with just one book or a more general problem ?. I Don't know how to inspect the source format to see what could have caused this. I used calibre 1.19 64 bit version.
I am just flagging this as something that may been more investigation, if anyone else has similr conversion experiences