I retired them mostly because the Ubiquiti management is much easier and that hardware also affordable (though the software is not open, so not a fit for your use case).
It's absolutely not all roses on the UBNT side of things. They are exhibiting some of classic signs of expanding too fast and stretching themselves a bit too thin. In particular their hardware lineup is starting to get overly broad and they aren't being aggressive about retiring older products and keeping the matrix simple, which of course in turn represents an increasing maintenance burden. And some of their hardware which was disruptively priced and fantastic value at launch is now getting very old in the tooth. The UniFi controller UI can be shallow for more then simple usage of things like DNS/DHCP/RADIUS, granted a lot of HN types may have their own separate appliances/servers for that. Their USG has always been a bit of an orphan and only recently has really started getting the serious attention it needs. They've got some features on high end hardware that while niche still haven't been fleshed out. Their EdgeRouter hardware is keeping up better though.
That said the update process has continued to be pleasant and solid, and their support even for old devices has been excellent. There are no required ties to any external services. The hardware itself has been very reliable, and even the RMA process for when something burned out on us was decent (2 minute wait to online chat on a Sunday morning and immediate RMA approval). They've been quite good on security updates for a number of the major issues that have come up over the last year, and have had no major snafus (that MikroTik one storing passwords as plain text was painful/disturbing to see). While enterprises will have more advanced needs for SoHo situations even if they're not open source I think UBNT is worth consideration, particularly for those wearing plenty of hats already who are ready to cut down on cognitive load a bit.
I've been running UniFi APs for years, but recently switched from my pfSense appliances to USG routers. I lost a lot of flexibility (especially for things like VPN configuration), but the simplicity and seamless management have been a huge time-saver.
I am puzzled by some of their new offerings—do they really expect any serious commercial customers to install lighting powered by PoE? Perhaps they're onto something innovative, but it seems like a distraction from their core business.
Yes, although I want to emphasize again that while they made a new hire specifically for the USG and it's seen dramatic improvements in the last year [1] it was still a kind of orphan child for a while and I still need to drop down to the shell sometimes for initial setup. Rock solid after that and simple and good integration with the overall site sure but it hasn't always been clear for someone starting from scratch how to get it up the first time in common SoHo situations. Also for those with gigabit links who want to run Suricata IDS/IPS (which requires turning off hardware offload), Ubiquiti just doesn't offer anything even remotely SoHo priced with the muscle for that right now. The low end USG "3P" (~$110) maxes out around 150 Mbps with IPS after the most recent update (an improvement from 85 Mbps before that) while the Pro (~$300) maxes out around 430-450. Only the XG can handle a gigabit or higher but that's $2500 and built with 8x 10G links, it's ludicrous overkill for those who don't want its other features and routing. Granted gigabit fiber links are far from the norm but they're gradually increasing and the HN crowd may be more likely to go for them then many, and the hardware in the USG 3P and USG Pro is just old now.
>I am puzzled by some of their new offerings—do they really expect any serious commercial customers to install lighting powered by PoE? Perhaps they're onto something innovative, but it seems like a distraction from their core business.
I definitely agree about distractions, though at the same time we should recognize that of course different divisions and people can be doing different things at the same time, development and engineering talent isn't necessarily fungible there. Still, they aren't a megacorp, overall resources and management bandwidth isn't unlimited either.
On the other hand at one point Ubiquiti had a real effort in the IOT space called MFi, but due to a lot of internal technical debt issues there (IIRC there) it essentially got canned, and they planned to eventually resurrect it on top of their more advanced foundations but haven't had the bandwidth. Maybe the lighting and their efforts to improve their security offerings are some first baby steps towards getting back into that? In fairness IOT has some of the same properties in terms of suckage that have made their networking efforts successful, and could also be a major market. Updates are often a pain or non-existent, the security story is awful, and much of it insists on using 3rd party cloud dependencies. There could be a real valuable hole there for Ubiquiti to fill were they to execute well enough, though I'd feel better about it if their core felt more tightly managed and foundations a bit steadier. We'll see I guess, and at least lighting should have been a pretty low R&D way experiment with it?
---
1: And for better or worse, the very fact of a piece of cheaper networking gear seeing years of support and improvements is depressingly unusual in the industry.
[1] DD-WRT struggled with 100Mbps to WAN - hardware NAT
(I'm not disputing your experience, but don't want other readers to conclude that 5GHz doesn't work on any of them.)