fdt_addr not setup by U-Boot/first stage bootloader - raspberry-pi3

Question

I'm working on setting up U-Boot for our systems using raspberry pi 3B plus and cm3. (Though atm I'm working just with cm3. I'm confident the same applies to the 3B plus)
I got rid of the default bootcmd run distro_bootcmd and load a first stage boot script first_boot.scr which then loads and sources the respected boot.scr.
%EDIT%
In my answer I mentioned already that the failure description is wrong. I just checked the fdt_addr variable at the wrong time, thus leading me on the wrong path.
%EDIT%
As we already have machines out in the field I'm working on updating them. Here lies the problem. If I setup a fresh device with the new architecture everything works just fine. If I update and older system though, fdt_addr is not setup, thus the boot failes.
I can of course manually setup fdt_addt, but I want to understand what could be the reason for fdt_addr not being setup. As far as I could find, the first stage bootloader is responsible. So I imagine something is not beeing triggered.
Updating in our case describes simply copying the new images to the device, moving u-boot.bin and first_boot.scr to the boot partition and then changing the config.txt to start U-Boot not the kernel directly.
While setting up a fresh machine is pretty much the same just with formatting the device and placing the files in the exact same places.
bootcmd: load mmc 0:1 ${scriptaddr} /first_boot.scr; source ${scriptaddr}
env load #<- %EDIT%
if test "${boot_count}" > 4; then
echo "###Loading factory kernel bootscript###"
setenv boot_kernel 2;
load mmc 0:2 ${scriptaddr} images/factory/boot.scr;
source ${scriptaddr}
elif test "${boot_count}" > 2; then
echo "###Loading factory kernel bootscript###"
setenv boot_kernel 1;
load mmc 0:2 ${scriptaddr} images/previous/boot.scr;
source ${scriptaddr}
else
echo "###Loading current kernel bootscript###"
setenv boot_kernel 0;
load mmc 0:2 ${scriptaddr} images/latest/boot.scr;
source ${scriptaddr}
fi
if test "${boot_kernel}" = 2; then
echo "###Loading factory kernel###"
load mmc 0:2 ${kernel_addr_r} images/factory/Image;
load mmc 0:2 ${fdt_addr} images/factory/device_tree.dtb;
elif test "${boot_kernel}" = 1; then
echo "###Loading previous kernel###"
load mmc 0:2 ${kernel_addr_r} images/previous/Image;
load mmc 0:2 ${fdt_addr} images/previous/device_tree.dtb;
else
echo "###Loading current kernel###"
load mmc 0:2 ${kernel_addr_r} images/latest/Image;
load mmc 0:2 ${fdt_addr} images/latest/device_tree.dtb;
fi
setenv bootargs console=ttyAMA0,115200 printk.devkmsg=on usbcore.autosuspend=-1 fsck.mode=force fsck.repair=yes
env save
booti ${kernel_addr_r} - ${fdt_addr}

Answer 1

The culprit was a line I sadly left out in my example as I deemed it unimportant...
I work with the boot_count variable also in user land, for that reason I save my environment to the drive so I can interact with it via fw_saveenv.
Because of that I need to load the environment at the beginning of first_boot.scr.
But if you try to load with env load when there is no uboot.env file alot of the environment information is lost. So now I do a check for uboot.env and save if it is not present and load if so.
I described the problem wrong in my question though. I check the presence of fdt_addr before sourching first_boot.scr at times, thus getting lead on the wrong track.
if test -e mmc 0:1 uboot.env; then
echo "Loading env"
env load
else
echo "Saving default environment to file"
env save
fi
if test "${boot_count}" > 4; then
echo "###Loading factory kernel bootscript###"
setenv boot_kernel 2;
load mmc 0:2 ${scriptaddr} images/factory/boot.scr;
source ${scriptaddr}
elif test "${boot_count}" > 2; then
echo "###Loading factory kernel bootscript###"
setenv boot_kernel 1;
load mmc 0:2 ${scriptaddr} images/previous/boot.scr;
source ${scriptaddr}
else
echo "###Loading current kernel bootscript###"
setenv boot_kernel 0;
load mmc 0:2 ${scriptaddr} images/latest/boot.scr;
source ${scriptaddr}
fi