Thanks @tannebil and @batwater for reporting on your setups and experiences over long periods in some detail. I'm encouraged that I've chosen UPB, and SAI as a vendor, for my power-line control needs. More thoughts follow.
@tannebil said: I love my UPB switches but will be replacing them at some point as they are not well-supported in the HA ecosystems that look to dominate going forward (Apple Homekit, Samsung SmartThings, Google Home, Amazon Echo). My UPB switches and HAI Omni have served me well but I think they belong to a fading era. People have built great DIY HA systems using them and I think a certain class of hobbyist will continue to do amazing stuff with them for many years to come but they are not going to be a mainstream consumer solution unless somebody releases a UPB hub with support for the emerging ecosystems.
Like X10 before it, I agree that UPB (and many other proprietary protocols across the HA space) is a niche product. HA itself is still a niche market within the total houseing/consumer market. Huge swaths of apartment dwellers & renters can't or don't invest in hardware upgrades for a home. As I see it there is little future in totally-wireless HA solutions, wireless is OK for *some* purposes *some* of the time but some features need copper to work reliably. Aside from ever-increasing RFI there is just the battery-power problem; the more devices one has the more frequent one will be hunting down & replacing batteries.
Though AmazonEcho, HomeKit, SmartThings, etc. are making inroads HA is still pretty immature. A lot of basic problems in the HA space are still "hard" and not yet solved. For example, presence detection alone, aside from all other data inputs -- not easy to implement with "5-nines" reliability. A "smart" home should not have to be told whether anyone is at home or not. But answering this question with 99.999% reliability, without any active input from a human being, isn't yet possible to do, not affordably.
For HA to really take off it needs become as ubiquitous and reliable as POTS (telephone) once was. One can buy an occupancy sensor for a room -- OK, maybe it's even a fancy dual-mode IR + microwave type unit that is 99% reliable. But by itself it is still not good enough to give an authoritative answer on whether someone is in a room. Say, for life-safety or security purposes. Would you trust an occupancy sensor alone to decide whether to arm an alarm system? What if someone is sleeping, what if the covers are thick & don't move. OK put strain gauges on the bed to see if someone's in bed. That'll work but it's not an off-the-shelf solution at all.
As a "hacker" type I'm definitely 2 standard deviations outside the mean, even within the HA space. I realized that to get what I want in the long run I -- a huge degree of interoperability & flexibility, as well as reliability and security -- I need to do it myself. In my case using OpenHAB. I looked at SmartThings and at MiCasaVerde & concluded that it is a bad joke to try to troubleshoot someone else's closed and often poor-quality firmware.
I think part of the future of HA may lie in highly advanced combined sensors with appropriate & configurable AI built into them. E.g. I would pay $200 or $300 for a single box the size of a tennis ball with a camera, mic, IR motion + microwave motion, + temperature + humidity + sizable CPU & firmware in it. That could be configured in multiple simultaneous "modes" for sensor feedback. E.g. the "occupancy sensor" mode would use AI routines to cross-verify sensor data before deciding whether someone was present. Was that motion a person or just a curtain blowing? The AI looks at the camera feed & detects not just motion in the frame but image-processes to see whether the object could conceivably be shaped like a human. To do this with 99.999% reliability outside of a laboratory setting is not an easy problem.
-----
Back to UPB and powerline control specifically, I think it is a particularly hard niche to be in because there's not *all that much* one can do with 120VAC power. It's not as "sexy" as Z-wave door locks & Amazon Echo. UPB does ON, OFF, DIM, TIMER, & grouped "scenes". By themselves these are useful, but not really a "killer app". And developing the reliable hardware to make them work is a lot harder than many low-voltage or wireless-only protocols. The 120V home wiring is inherently high-power and often full of potentially brutal noise, i.e. motor brush commutators, gigantic pulses from compressor cycling, continuous chop from fluorescent ballasts, etc. So at the engineering level it's hard to begin with, and it's made harder still by much stiffer (as they should be) regulatory requirements and certifications, everything needed to ensure that a UPB device won't shock a user, short out entirely, or cause a building fire.
I'd never have considered buying $80 to $130 light switch replacements for the stand-alone features of UPB. But integrated with a larger HA (OpenHAB as I mentioned,) the case becomes compelling. When I can have my UPB lights go on / off / dim not just on a timer, but according to whether I am at home, what season it is, and whether it's cloudy outside or not, that brings my lighting closer to a seamless integration with the way I live. Aside from lighting of course there is integration with any conceivable 120V device, e.g. HVAC equipment, pumps, etc. To be able to control it all with very high reliability, safely, *and* without having to run separate, cost-prohibitive and often physically-vulnerable low-voltage communication cable runs makes UPB compelling.
I want my HA system to be as reliable as my hot & cold running water and mechanical door locks are. I don't often think about those technologies, they "just work" because they have been engineered with high reliability to as a design criterion. I think a carefully-thought-out HA system that does not rely on cloud services, or even an internet connection, that has no wireless in any critical communications path, and that is built on reliable hardware can approach the "hands off" performance I seek.