I have managed to use your script and end-up with two part circles for those which overlap the 180 longitude line.

The way I did it:

• Create your buffer layer

• Reproject it to EPSG:3832 ("Reprojected" layer in my screenshot)

• Create a layer(4326) with a line at 180 deg longitude

• Reproject that line to EPSG:3832

• Buffer that line (still in 3832) as a "BufferedLine" layer. Buffering width should be wide enough to be sure that it spans across the 180 line

• Set the current rendering CRS to 3832 to have a clearer view, you should see the circles across the antimeridian, visually split by the buffered line:

• Apply a Difference transformation (Input: Reprojected, Overlay: BufferedLine, set the Input options to "Do not filter" for Invalid Features if you have errors. This should split the cirle in two parts that don't overlap the 180 longitude line. I don't know why, at first it did not work, the circle was not split, I had to try again.

• Resulting "Difference" layer should show fine in EPSG:4326.

• You can reproject "Difference" to EPSG:4326

Easter side, faulty geometry in green, new half geometry in grey:

Western side:

Full view without the inappropriate geometry:

I hope this works for your whole dataset

I have managed to use your script and end-up with two part circles for those which overlap the 180 longitude line.

The way I did it:

• Create your buffer layer

• Reproject it to EPSG:3832 ("Reprojected" layer in my screenshot)

• Create a layer(4326) with a line at 180 deg longitude

• Reproject that line to EPSG:3832

• Buffer that line (still in 3832) as a "BufferedLine" layer. Buffering width should be wide enough to be sure that it spans across the 180 line

• Set the current rendering CRS to 3832 to have a clearer view, you should see the circles across the antimeridian, visually split by the buffered line:

• Apply a Difference transformation (Input: Reprojected, Overlay: BufferedLine, set the Input options to "Do not filter" for Invalid Features if you have errors. This should split the cirle in two parts that don't overlap the 180 longitude line. I don't know why, at first it did not work, the circle was not split, I had to try again.

• Resulting "Difference" layer should show fine in EPSG:4326.

• You can reproject "Difference" to EPSG:4326

Eastern side, faulty geometry in green, new half geometry in grey:

Western side:

Full view without the inappropriate geometry:

I hope this works for your whole dataset

I have managed to use your script and end-up with two part circles for those which overlap the 180 longitude line.

The way I did it:

• Create your buffer layer
• Reproject it to EPSG:3832 ("Reprojected" layer in my screenshot)
• Create a layer(4326) with a line at 180 deg longitude
• Reproject that line to EPSG:3832
• Buffer that line (still in 3832) as a "BufferedLine" layer. Buffering width should be wide enough to be sure that it spans across the 180 line
• Set the current rendering CRS to 3832 to have a clearer view, you should see the circles across the antimeridian, visually split by the buffered line.
• Apply a Difference transformation (Input: Reprojected, Overlay: BufferedLine, set the Input options to "Do not filter" for Invalid Features if you have errors
• Resulting Difference layer should show fine in EPSG:4326.
• You can reproject Difference to 4326

Easter side, faulty geometry in green, new half geometry in grey:

Western side:

Full view without the inappropriate geometry:

I hope this works for your whole dataset

I have managed to use your script and end-up with two part circles for those which overlap the 180 longitude line.

The way I did it:

• Create your buffer layer

• Reproject it to EPSG:3832 ("Reprojected" layer in my screenshot)

• Create a layer(4326) with a line at 180 deg longitude

• Reproject that line to EPSG:3832

• Buffer that line (still in 3832) as a "BufferedLine" layer. Buffering width should be wide enough to be sure that it spans across the 180 line

• Set the current rendering CRS to 3832 to have a clearer view, you should see the circles across the antimeridian, visually split by the buffered line:

• Apply a Difference transformation (Input: Reprojected, Overlay: BufferedLine, set the Input options to "Do not filter" for Invalid Features if you have errors. This should split the cirle in two parts that don't overlap the 180 longitude line. I don't know why, at first it did not work, the circle was not split, I had to try again.

• Resulting "Difference" layer should show fine in EPSG:4326.

• You can reproject "Difference" to EPSG:4326

Easter side, faulty geometry in green, new half geometry in grey:

Western side:

Full view without the inappropriate geometry:

I hope this works for your whole dataset