TL-WR1043ND v1 JFFS2 errors

After flashing 19.07.2 (ath79) and issuing some "reboot" commands from CLI (serial connection) JFFS's health is getting bad:

jffs2_build_xattr_subsystem: complete building xattr subsystem, 0 of xdatum (0 unchecked, 0 orphan) and 0 of xref (0 dead, 0 orphan) found.

jffs2_build_xattr_subsystem: complete building xattr subsystem, 3 of xdatum (2 unchecked, 1 orphan) and 3 of xref (1 dead, 0 orphan) found.

jffs2_build_xattr_subsystem: complete building xattr subsystem, 4 of xdatum (2 unchecked, 2 orphan) and 6 of xref (2 dead, 0 orphan) found.

Those counters increase at every boot.
This is reproducible on my unit even after a jffs2reset. Is anyone else experiencing the same?

The issue is still present after installing the 19.07.3 factory release (starting over).

I see similar behaviour (orphan and dead incrementing by 1 after reboot) with a mr3020 v3.2 (this is ramips not ath79).

If I reflash it the number or orphans and dead goes back to 0 but reboots increase them by 1 again. To be honest it's a weird behaviour especially that it goes down to 0 after a flash...

At least with my mr3020 it happened with the latest 19.07 self-compiled (nothing fancy there, just support for android tethering and sqm added, stuff that shouldn't affect jff2).

On the other hand the 841nd v5 and v9 with latest 18.06 self compiled don't experience such behaviour. They all show 0 orphan/dead after reboots.

by searching in bug reports and on this forum i do see other numbers than 0 for orphan/dead for other people too in the log posted

@dash17291 can you revert to 18.06 (maybe wait for 18.06.9 to be released) and see if this weird behaviour stops? (i'm not that lucky 3020 v3 support starts with 19.07...) (you gonna have to not keep the settings and redo them if u downgrade to 18.06 cause u gonna move from ath79 to ar71xx)

L.E.: By looking at the logs on the forum this behaviour looks to happen with kernel 4.14 (no clue if newer kernels are affected also). It also doesn't look to be a platform target specific thing.
It might happen in 5.4 also, not that many logs with 5.4 kernel cause that's master.
The common thing looks to be kernel 4.14 or newer. 4.9 doesn't look to be affected.
(Even if I backport support for my mr3020 to openwrt 18.06 it will change nothing cause in 18.06 ramips is on 4.14 kernel... Too lazy to read all the 4.14 changelogs to see if it's something that broke in 4.14, can easily be before 4.14 but after 4.9...)