As I suspected, the free space usage does not decrease with a reboot, meaning that the written-to space is not in /proc or /tmp. This is very strange then, since the find results indicate that no file is being written to rapidly, but the block_dump indicates that some blocks (not inodes) are being written to.
It seems to point to a kernel bug that spams WRITEs to contiguous free blocks but not storing any meaningful files to those locations.
As I suspected, the free space usage does not decrease with a reboot, meaning that the written-to space is not in /proc or /tmp. This is very strange then, since the find results indicate that no file is being written to rapidly, but the block_dump indicates that some blocks (not inodes) are being written to.
It seems to point to a kernel bug that spams WRITEs to contiguous free blocks but not storing any meaningful files to those locations.