Blocky corrupted image pattern for Sentinel 2?

When plotting the data, I noticed that some of the S2 images have a strange blocky pattern. I have checked that I am downloading the data correctly using the checksum, double checked the image in QGIS, and checked that the CLP band is not 255 (ie. the noise value).

Is this just noise that is not explicitly represented by the CLIP band?

The plot below has the months in rows and bands in columns and represents chip 28ac85b5. I noticed this pattern in 3 out of the 100 chips I plotted.