Would you be able to add a test, that fails without your change and passes with it?

Ah, only internal TiffImagePlugin.ImageFileDirectory_v2 is ever used to parse the TIFF IFDs (in both Image.open() and TiffImagePlugin._load_libtiff(), there's no way to actually see how libtiff is re-interpreting the IFD as far as I can tell, so not really sure how to test this...

https://linux.die.net/man/3/tifffdopen

The default behaviour, for backwards compatibility, is to enable strip chopping.

Is it possible that this PR doesn't actually do anything?

I only know this: https://gitlab.com/libtiff/libtiff/-/blob/29219a5bd14040d60e376324a52ef91b68808143/libtiff/tif_open.c#L254-257

But it does seem the flags are initialized by default with the intention to do the chopping: https://gitlab.com/libtiff/libtiff/-/blob/29219a5bd14040d60e376324a52ef91b68808143/libtiff/tif_open.c#L161-164

However, there was also recent (not yet released) PR that seems to indicate the default wasn't working on libtiff CMake builds (which is what Pillow uses lately AFAIK): https://gitlab.com/libtiff/libtiff/-/merge_requests/271

If this is the behaviour Pillow was expecting, doesn't hurt to be explicit ;)

Ok, if the default has been wrong in previous libtiff versions, then it seems like this would be good, as it will fix that.

libtiff 4.4.0 has now been released, fixing the default behaviour.