There is a fairly long development cycle with these things, so you will not always see the latest. I don't believe Iray was actually chosen to attract DAZ users as to RL's motivation, as always, we can just speculate.
Yes, Eevee is amazing, and it gets better real fast, and like I said, I understand that they want to attract Daz users, but damn, I ray really sucks, even the Daz users bitch it on their forum, it is slow compared to Cycles, Pro Render, and pretty much every render engines out there, DXR is also a big thing, definitely a game changer for game developers and VFX artists as well, to implement Iray when there are way better solutions just doesn't make sense to me, but as you said, maybe the new api will make it easier for developers to integrate new render engines, lets hope that they will take the bate! They are diversifying their target users between 3d hobbyists and game designers, and that is going to be a smart move for the long haul. But now I am back in, and using both tools. To be honest, I had stiopped using iclone as soon as iRay came out for daz. To offer a way into iclone from those platforms is a guaranteed money maker. That said, Daz/Poser is the biggest base of money-spending 3d users on the planet. And it also runs on both nvidia and amd and CPU only or a mix of all three.
The built in Eevee renderer is amazing, and while not technically real time, it gives 4k render results in seconds per frame. I myself am hoping for a one button convert to Blender solution. Hopefully the python api is going to result in a whole slew of converters to other render engines. You are technically right.there are other renderers out there. EDIT: Just thought of something else important and that is support for particles, such as PopcornFX, and things like water, which is currently absent in Iray, but will hopefully be added. Looks like this will not help sell AMD cards. Anyway, I was wondering how Pro-Render would actually do on Nvidia GPUs (as I am not willing to switch to AMD), and the result was quite surprising, judging from this test. I guess many of us have our favorites, so it would be quite a task for RL to make us all happy. It has always been the plan that multiple renderers would be supported, so the hopefully the Python API can help with this. Its value is also determined by how well it supports animation. The relative speed between renderers can only be determined in an actual environment where the same project is rendered. Also, one question to ask is how easy the interfacing is between iClone and a specific renderer, especially as far as materials go.