Performance: why is PdfWriter(clone_from=...) much slower than PdfWriter().append_pages_from_reader(PdfReader(...)) ?

Context: I'm trying to replace my uses of pdfrw by pypdf in various scripts that I have,
and pypdf is still revealing to be quite slower...

In some case, I'm witnessing a factor x5 speed difference between using PdfWriter(clone_from=...) and PdfWriter().append_pages_from_reader(PdfReader(...))

Environment

$ python -m platform
Linux-5.10.102.1-microsoft-standard-WSL2-x86_64-with-glibc2.29

$ python -c "import pypdf;print(pypdf._debug_versions)"
pypdf==3.15.4, crypt_provider=('cryptography', '39.0.2'), PIL=9.4.0

Code + PDF

This is a minimal, complete example that shows the issue:

from pypdf import PdfWriter
PdfWriter(clone_from="2020-LaCaravelle_LaMortPrendDesVacances.pdf")

Versus:

from pypdf import PdfReader, PdfWriter
writer = PdfWriter()
writer.append_pages_from_reader(PdfReader("2020-LaCaravelle_LaMortPrendDesVacances.pdf"))

Source PDF file: https://chezsoi.org/lucas/2020-LaCaravelle_LaMortPrendDesVacances.pdf (3.9Mib - 148 pages)

Execution times

$ time ./repro_writer_fast.py $jdr/2020-LaCaravelle_LaMortPrendDesVacances.pdf
real    0m0,819s

$ time ./repro_writer_slow.py $jdr/2020-LaCaravelle_LaMortPrendDesVacances.pdf
real    0m4,551s

Start by reproducing the comparison between PdfWriter(clone_from=...) and PdfWriter().append_pages_from_reader(PdfReader(...)) using the supplied 148-page PDF and the minimal examples. Profile or inspect both entry points to identify the source of the timing difference, then verify that equivalent cloning no longer has the reported performance gap.