It has sort of been that way in the past, so I wouldn't be surprised if there were similar issues with Ryzen at first.
However, I expect that there will eventually be some workstation grade motherboards from vendors like Supermicro which will be more direct about guaranteeing ECC support. This will be especially true if AMD decides to release some “official” small server grade CPUs on the AM4 socket (kind of like Intel’s Xeon E3 which uses the regular consumer socket, but requires a different chipset). I don’t know if there are any plans for such a CPU from AMD, but it wouldn’t surprise me considering there’s probably some real demand for something a little less expensive and powerful than the 32C/64T Naples server CPU. Rebranding a consumer 8C/16T Ryzen seems like a logical step, with the only cost being validation.
The thing to remember is that supporting ECC on an AM4 motherboard basically amounts to running a few extra traces between the CPU and DRAM slots, and adding support in the BIOS to initialize the memory controller properly. My experiencing in implementing boot code to initialize memory controllers leads me to believe that it’s probably one or two days worth of work to actually get it working (once everything else is done).
The real cost comes from testing and validation, which is why some manufacturers are so hesitant to give a straight answer regarding ECC support on consumer platforms. There is no point in spending the money to validate a feature that the majority of their customers think is a complete waste.
By offering “working but not validated” ECC support, they’re able to satisfy the demand of those who want to use these chips for small non critical servers without undercutting their higher margin server grade components.