We run Triton and individual SmartOS boxes in production. They have pretty much replaced all of our VMWare and linux-based hypervisors save some very specific use cases, mostly relating to GPU passthrough and NVIDIA.
The case with OpenZFS does worry me as well. I fear the developers start slowly introducing linuxisms thus sacrificing portability and stability for the great penguin.
Generally there hasn't been anything game breaking and our biggest issues have been with running out of logging space for the core services (went with too low capacity disks in the beginning, doh) and one failed upgrade which didn't even take down the whole cluster during fixing. Our clients can now also provision their own VMs with the Triton API unlike with VMware which required admins to do it. Bhyve like KVM before has also been rock solid in everything we run from simple web servers to kubernetes clusters.
For issues we've found the Joyent/SmartOS IRC channels to be excellent and they have helped us tremendously in debugging and fixing things. It's the best support I've encountered for a FOSS product by far and one of the biggest reasons I'm such an illumos advocate now.
Not running Manta at the moment but it's on the menu. The demand for object storage hasn't been high enough to justify the spent time to set it up for now.
Not sure about linuxCN either. I would rather run !linux on bare metal as much as possible, and having purged most of the penguins from infrastructure it would feel a bit weird to go immediately back ;)
In addition, we also have SPARC hardware running on OpenBSD due great hardware support, including ldoms! Of course would be nice to run illumos on them too, but alas...
The case with OpenZFS does worry me as well. I fear the developers start slowly introducing linuxisms thus sacrificing portability and stability for the great penguin.