I tried running my desktop from a 120GB+1TB writeback cache setup, but every time the system is rebooted the entire cache is marked dirty, and takes several hours to flush (it does complete).

Two different setups were tested (below), the issue only affected the root mount point, not home.

Any ideas on how to debug/fix this?

First setup (encryption on top of LVM):
    One GPT partition on each drive, used as PVs.
    A single cached LV.
    / is an EXT4 partition, encrypted with LUKS.
    /boot and /boot/efi are simple partitions on the slow drive.

Second setup (LVM on top of encryption):
    A couple LUKS volumes, one on each drive, serving as PVs.
    One VG backed by the fast and slow LUKS volumes.
    / on a cached LV.
    /home on a separate cached LV.
    /boot and /boot/efi are simple partitions on the slow drive.

System:
    Fedora 30, KDE spin.
    Kernel 5.1.7-300.fc30.x86_64

--
Lucas Vinicius Hartmann