I can submit a new PR if you'd like.

Thanks for the PR! I'll take a look tomorrow. In the meantime, can you explain a bit on what things didn't work before? Either edited in the PR description above, or a new issue?

I can submit a new PR if you'd like.

How familiar are you with git? You should be able to just commit to the main branch of your fork where you did this work to update this PR.

Why don't I submit a bug report for the two things that were changed.

As for Git, I'm somewhat proficient but I'm not hugely familiar with how Github handles things. I'll try pushing an update to my repo and see if it propagates here.

I'll be submitting a second bug report shortly regarding the changes to to callback function. In a nutshell it's a workaround for what appears to be a Python bug (likely system-dependent).

All of the changes that we've been discussing plus the fix for #4 have been implemented and pushed to Github. For installing the callback I used my implementation (a lambda); as this is your project though it's ultimately your decision as to which method to use for this. Currently I'm passively testing the code (it's being triggered every minute or so) and I haven't seen any issues so far.

The most recent push (14ec5f8) covers everything that we've been discussing. This code has been tested for the past 24 hours with multiple file-backed layers being watched and no issues have been seen. In my opinion this code should be ready to be merged.

Thanks for the finding the bugs, reporting and fixing them! I will make some minor changes and make a release this weekend.

Thanks for the finding the bugs, reporting and fixing them!

¡De nada! And thank you for writing such a useful plugin!

I will make some minor changes

There's always something missed: lines 66 and 67 aren't needed and can be deleted (that's the from qgis.core import QgsProject statement and its comment). I've updated my code and am passively testing to ensure it can be safely removed (which it should as it was only used for retrieving the layer by its ID in my initial push). This change was pushed to my copy of the repo after you merged the code (4e49c7a).

I've removed some lines, ran an autoformatter, added a test and some more in 1577f35. Most impactful, I've shortened a check you did for re-adding the watch at 1577f35#diff-859e6c5392dbec182d14bd578888ecdd0c6c29b9007f5abce8d7654f541b2418R356. Can you confirm it still works as it should?

edit: See, it was good to check, I think I inverted the condition there.

Can you confirm it still works as it should?

I read through the diff and saw a few things, mostly minor (stylistic), one or two that might be issues. I'll pull the code and test it then annotate the aforementioned diff.

Correction: I'll annotate the diff between your merge of my pull request (5083128) and your most recent push (1f35f16). [edit: Or not, Github won't let me annotate that diff.]

How are you running test.py?

I added comments to 1f35f16 and 1577f35

The current code is running fine on my system (it's been running with watched file updates [for multiple files] every minute or two for the past 4 hours).

Great, thanks for testing! I'll make a release tomorrow.

There are a few suggested changes in the comments I made.

Amongst other things, can you test this on Windows or would you like me to install QGIS in a VM and test it there? [edit: I did the latter.] To be specific, on Windows / characters in file:// URIs may need to be changed to \.

I was right, file:// URI aren't handled correctly on Windows (there's even an odd added quirk). Fixing it looks trivial. I'll submit a bug report and a new PR to fix the issue.