Well yes, but actually no
From a software standpoint, I personally think the polish for Bluetooth location trackers has reached a point that it can be considered mature enough for "everyday" use. I personally don't have a ton of experience with them, but from what I have done and do know, they're in a good place.
The real problem is that all of these Bluetooth trackers use what's known as Received Signal Strength Indicator (RSSI.) Basically what this does is measure the signal strength and (along with some fancy things), uses the basic assumption that signal strength falls off as a function of distance to approximate the distance away. This is why an unstable (or too stable) connection or random signal drops due to interference leads to poor performance. This is inherent to the way that these trackers works.
There's a better way of tracking though: Time of Flight (ToF.) Since signals travel at a known speed (~the speed of light), this property can be used to accurately track distance, independent of signal strength (and therefore independent of any interference.) This is the way that GPS, 802.11v, and many other technologies work. If you lose your Apple Watch (if it's not GPS equipped), it can relatively accurately locate itself.
However, Bluetooth is fundamentally incapable of accurately using this technology. The problem is the Bluetooth bandwidth. Bluetooth operates between 2.402 GHz and 2.4800 GHz, (plus a 2 MHz bottom and 3.5 MHz top guard band), split into 79 channels of 1 MHz each. Speed of light / 1 MHz ≈ 300m (it's slightly more complicated than this, but close enough for our purposes.) This is so ridiculously inaccurate as to be completely impractical. Of course, it is possible to do better by averaging multiple runs and some of the newer features of Bluetooth, as done in this paper, but the fundamental restrictions of Bluetooth remain in place. With infinite time, would the law of large numbers end up getting you pretty good data? Sure. But at a more reasonable number of samples? Well, it's not really any better than RSSI. And at that point, why bother? Getting the RSSI value is so trivially easy that a middle schooler can (and has) done it. And if you're creating a product, how many people are really going to go through the trouble of, say, posting a question on a hardware recommendation forum? How many people who read an Amazon listing will know the difference between RSSI and ToF, much less care?
It's not like good location technology is out of reach anyways though. Like I said before, the Apple Watch simply uses more bandwidth, and thus can locate itself more accurately. It's not really a hard problem to solve from a technical standpoint. The issue is one of adoption: we've all got convenient little Bluetooth gadgets, everyone knows what they are, the market is small, how much do people really want to push for a better (consumer grade, at least) locator beacon?
Thus we find ourselves in the current state of affairs. Yes, they're polished, finished, marketable, and actually sometimes quite useful products. There is value to having them—especially if you just want to know if you're even in the same room as your wallet or keys. They're just not the magical locator beacon you might want.
Anyways, so glad someone asked this question! I did some work on this a number of years ago, and it feels good to share it to wider audience. It's also good to see new research is still coming out and pushing the boundaries of what we thought was possible.
