RAID1 two chunks of the same data on the same physical disk, one file keeps being corrupted

[ein <ein.net@gmail.com>]

Dear devs and users,

I used BRTFS for few months in RAID1 mode on Debian 12 (6.1.0-21-amd64 #1 SMP PREEMPT_DYNAMIC Debian 
6.1.90-1 (2024-05-03) x86_64 GNU/Linux, btrfs-progs v6.2).
I created my filesystem by issuing:
mkfs.btrfs -d raid1 -m raid1 /dev/sda1 /dev/sde1 /dev/sdf1

Those are 2TB WD Reds, mix of CMRs and SMRs with good S.M.A.R.T. stats.
I am using on-die-ecc RAM memory modules.
I never did balance or replace any device.
I had couple of unexpected hangs because of nvme power management which made my root fs unavailable, 
but hopefully it's been fixed by installing new firmware for WD black nvme.

How it's possible that btrfs kept same chunk of data on the same physical device?

Jun 02 23:27:54 node0 kernel: BTRFS warning (device sdf1): csum failed root 256 ino 259 off 
140290392064 csum 0x1315675d expected csum 0x49271c1b mirror 2
Jun02 23:27:54 node0 kernel: BTRFS warning (device sdf1): csum failed root 256 ino 259 off 
140290392064 csum 0x1315675d expected csum 0x49271c1b mirror 1

The corrupted file is qocw2 image with Windows 7 on it and I think I am able to corrupt this file 
and only this file on daily basis.
I resorted my filesystem from backup, by:
1. wipfs any singatires on my hdds,
2. recreating fs from scratch,
3. coping over new data, few days later I see the same issues:

Jun10 07:22:14 node0 kernel: BTRFS info (device dm-10): read error corrected: ino 259 off 
39193079808 (dev /dev/mapper/vg1-lv1 sector 670864280)
Jun10 07:22:14 node0 kernel: BTRFS info (device dm-10): read error corrected: ino 259 off 
33579532288 (dev /dev/mapper/vg1-lv1 sector 199031056)

I don't think that it's RAM related because,
- HW is new, RAM is good quality and I did mem. check couple months ago,
- it affects only one file, I have other much busier VMs, that one mostly stays idle,
- other OS operations seems to be working perfect for months.

Sincerely,