For me the most important aspect of ECC is not making the memory reliable. It's telling me the RAM is becoming unreliable, which you know because the number of errors detected and corrected is rising. DDR5 ECC does not provide that information.
Once you start seeing ECC RAM errors you know it might be the cause of system failures you see. Knowing a specific piece of hardware might be failing is far better than a random crash that might be caused by memory, or the CPU, or a software bug, or PCI bus failure and a zillion other things.
HDD SMART is the same thing. Once you see errors from a drive rising you know it won't be long before it fails. One of the negatives of SSD's is you don't get the information, so they fail without warning.
That aside, DDR5 ECC is not end to end. It might not be the RAM calls that are failing. It might be getting the bit to the pins on the chip, or the connector, or the memory bus, or the CPU RAM interface. CPU ECC covers all of that.
