Google Wifi System 3-pack (Open Box)

I thought a Google Wifi mesh would be ideal for a property with several buildings in rural Maine, allowing access to our limited DSL-based internet from various locations within a small campus. It could, I assumed, exploit existing cat5e wired connections where they exist, and fill in via wireless mesh connections elsewhere. I'd read rave reviews, and had great expectations based on Google's reputation, other products, and various contacts there. So I bought a 3-pack of "Wifi Points" and set up a little trial network on a corner of the property. I was alarmed to discover that I couldn't expand the mesh by adding a WP whose only connection to the primary WP was via wired ethernet, but eventually discovered the workaround of configuring via nearby wireless connection and then moving the new WP to its final -- wired -- location, and got a 3-node network working. After exercising it a bit for a few hours, I (optimistically, it turns out) ordered a bunch more WPs with the idea of covering the property with a uniform mesh. That unrealistic goal cost me more than a kilobuck and many, many days of frustrating configuration hell. I won't detail the myriad problems I ran into here, although I have many email exchanges with Google customer service that document most of them. But some of the high points: (1) Configuration requires an IOS or Android app. I was using IOS, and found the app VERY unsatisfactory. On many occasions it just hangs; in other case, it fails without giving any useful explanation. My most recent complaint is that it complains that the mesh is "offline", and won't do anything with it, despite a valid connection to the mesh network by the ipad running it. (2) Virtually zero diagnostic information is available. The mesh configures itself by exploring the connectivity graph among WPs, and building an acyclic spanning tree for routing packets; but it hides the structure of this routing configuration (apparently from the customer service people as well as the user). If the app is working (not always the case), you can query connectivity of devices (like computers and ipads), and determine which WP they are connected to as well as their recent bandwidth usage -- good features. But there's apparently no way to discover the connectivity among WPs, and some evidence that it is not always sensible. (3) Misrepresentation of status. On many occasions I've built configurations of 6 or 7 nodes that the app thinks are working "great", with white lights on each node (signifying no problems). But allegedly working WPs were radio-silent: after testing over long periods with many devices, supposedly working WPs had NO discernible wifi activity (eg, not broadcasting the SSID). I've observed this behavior with many of the 12 WPs I now own, and am convinced its not a hardware failure (despite the assurance by some Google customer service people that this "can't happen" with working WPs). (4) Handing out bad IP addresses. On several occasions, I had an otherwise-working mesh with a new WP plugged in, waiting to be set up, and observed the following bizarre behavior: a device connects to an established WP (nowhere near the node waiting to be set up), and gets assigned an IP address of the form 192.168.84.* rather than expected 192.168.86.* addresses Google has chosen for their mesh. I turns out that the setup-mode device advertises another SSID on .84.* addresses for temporary connections during setup, a reasonable (and commonly used) approach. But the fact that the mesh gets confused by this extra Google-supplied DHCP server and uses it (rather than the DHCP of the primary WP) seems remarkable. This bug is avoidable by not trying to use the mesh while a WP is in setup mode, but seems one of many examples of extremely sloppy design. (5) Self-wounding. In the interests of making the mesh self-healing, the routing configuration is periodically re-evaluated and perhaps changed. However, on several occasions I went to bed with a working (or mostly working -- perhaps a node or two "offline") and woke up to a completely non-functional mesh -- red lights everywhere, no network connectivity. Of course, I had no clue as to what happened: no log, no error report, no diagnostic. Just sudden death. There are many other issues that can be added to this list; these are just representative examples. About customer service: I talked to a number of gwifi service people, who were uniformly pleasant and polite (despite growing frustration on my part). Each was armed with a set of steps and suggestions for chasing the simplest of bugs, as well as some pieces of inconsistent and/or bad advice. One insisted that I replace all my cheap/dumb gigabit switches with managed switches. Several suggested that the switches might be blocking critical packets in their spanning tree protocol. Some insisted that using LAN vs WAN ports of secondaries is critical, others claimed it didn't matter. Generally, when I heard suspicious advice, I asked to speak to a supervisor (or more expert colleague); while this may have irritated some agents, they generally complied politely. I eventually wound up communicating (mostly by email) with a competent technical guy who seemed well intentioned but simply didn't have the resources to solve the problems I was presenting him with. He had the ear of a "product expert" (who I would have loved to talk to directly), but we apparently exhausted this pocket of expertise as well. I've been forced to conclude that there are serious flaws in the GWifi system that will require substantial additional development effort to correct. If I put my dozen WPs in a box in the attic for a year, its possible that some future firmware upgrade will make them work; however, by then I hope to have installed a mesh technology that actually works. Having stupidly discarded the Apple-esque packaging they were shipped in, my WPs seem destined to be distributed as party favors for some geek gathering or wind up in a landfill somewhere. Its worth noting that other have had similar frustration with this product. If you dig deeper, looking e.g. at discussion in technical forms, you get a different, very mixed picture from what is portrayed in the superficial reviews. If you're considering this product, I suggest you do this additional research; I wish I had done so.