Comment # 21 on bug 1198668 from Ulrich Windl

systemd-analyze blame said (Leap 15.4):
3min 46.428s plymouth-quit-wait.service
3min 26.981s purge-kernels.service
...
Startup finished in 813ms (kernel) + 15.409s (initrd) + 4min 7.745s (userspace)
= 4min 23.968s 
graphical.target reached after 4min 7.705s in userspace

However there are two things I don't understand, considering the unit file:
# /usr/lib/systemd/system/purge-kernels.service
[Unit]
Description=Purge old kernels
After=local-fs.target
ConditionPathExists=/boot/do_purge_kernels
ConditionPathIsReadWrite=/

[Service]
Type=oneshot
Nice=19
IOSchedulingClass=idle
Environment=ZYPP_LOCK_TIMEOUT=-1
ExecStart=/usr/bin/zypper -n purge-kernels
ExecStartPost=/bin/rm -f /boot/do_purge_kernels

[Install]
WantedBy=multi-user.target
------------
First which unit is waiting for this one to finish, and if waiting is intended,
it makes little sense to use "Nice=19; IOSchedulingClass=idle" as this just
makes the unit slower.

I found it:
/usr/lib/systemd/system/YaST2-Firstboot.service:After=purge-kernels.service
/usr/lib/systemd/system/YaST2-Second-Stage.service:After=purge-kernels.service

One comment says:
# Run after kernels are purged to prevent a zypper lock (bsc#1196431)
After=purge-kernels.service

Well, but even if that unit has nothing to do it will delay the whole process
in case there is something to do.