It would be a strange coincidence as I have shown, there is a difference for the 2-letter subfolders between unsorted order presentation and processing them, which is sorted in some way.
Of course I can only observe and this is what I am seeing and I have never observed it different than that.
What could that mean? Maybe that there is no significant performance impact for sorting it as suggested. But for sure my suggestion if implemented would not change anything for the bad for Windows nodes.
I cannot say how much effort would be required to implement my suggestion. Only analogy I have is a database like MySQL or something. If it is similar trivial like adding an ORDER BY clause to a MySQL statement then I don’t see much time and effort wasted.
I am not a coder, I don’t know anything about BTRFS and NTFS and if they have features that could be utilized for improvement.
Judge for yourself. I am seeing similar situation on several nodes. This is just one example:
ls -1 /trash/ukfu6bhbboxilvt7jrwlqk7y2tapb5d2r2tsmj2sjxvw5qaaaaaa
2024-06-19
2024-06-25
2024-06-28
2024-07-04
2024-07-06
2024-07-07
2024-07-08
2024-07-14
2024-07-20
2024-07-23
2024-08-05
2024-08-06
2024-08-16
2024-08-17
2024-08-25
2024-08-26
2024-09-02
And this is the order they supposedly will get deleted (if ever):
ls -1U /trash/ukfu6bhbboxilvt7jrwlqk7y2tapb5d2r2tsmj2sjxvw5qaaaaaa
2024-08-17
2024-08-25
2024-08-16
2024-08-26
2024-07-06
2024-07-20
2024-06-19
2024-08-06
2024-06-25
2024-07-14
2024-08-05
2024-07-23
2024-09-02
2024-07-08
2024-06-28
2024-07-04
2024-07-07
Why “if ever”? Because if you leave it to the file system/OS there is no guarantee that it will not place a more recent folder before older ones as you can see where folder 2024-09-02 is in line before 2024-07-07.
With my suggestion, there would be guarantee that old folders will be deleted as they would be always the first ones that the deletion process would be working on. The oldest date folder on the node above is approaching 3 months old, which is not what it could be.