In mid-November I ordered one of these for a small-office client to replace a Linksys/Cisco SRW2008 which the office had outgrown. I chose the TEG-S16Dg based on the fact that the office didn't need a managed switch, its excellent price-per-port when compared to other unmanaged switches, and its consistently-positive reviews both here and elsewhere. In this office there are 10 users backed by a Real Server, with consistent high-bandwidth utilization (lots of large document and image manipulation and cataloging). Given that the switch is unmanaged, installation is indeed plug-and-play. My observations were that the performance was on-par with the Cisco, and there were no issues with either Auto or hard-coding speed and duplex on the server, network peripherals, and client machines. All in all, a great box at a great price.
And then the problem started.
One month after installation, after running 24x7 with no issues, the switch just stopped forwarding between certain ports. The links didn't actually go down and there were no other indications of a failure, but random devices just couldn't connect any longer to random other devices. Bounce the switch and all was well for about 10 minutes, and then the problem started again. I say "random" above because there didn't seem to be any consistency about what ports would be affected, except in the case of the server - in all cases, that port would stop forwarding (whichever port it happened to be - I did try a number of different ports) - but otherwise, random devices among all 16 ports would be affected (not even in a particular port bank, which is a common failure mode), and the next failure after reboot would affect different ports, which ports would remain affected until the next boot. An investigation of the server revealed no warnings or issues, and I concluded with changing both the NIC and the cables just for giggles. No effect - same problem.
I order from Amazon primarily because of Prime, their wonderful return policy, and excellent support reps. I received a same-model replacement right away, installed it, sent the bad one back, and, after a couple weeks of solid performance, chalked it up to a one-off bad unit. No such luck. The exact same issue has now occurred with the new unit.
I called TrendNet support. They claim no knowledge or reports of this issue, and given that the switch is unmanaged, there's no user-serviceable action that can be taken, such as a firmware upgrade. Googling also offers no reports of such a problem.
Given that lack of reports, and given the unlikelihood that I've received two bad switches with identical failure modes, my conclusion is that there is a design flaw in the product that can't handle some event that's happening in this environment (though I can correlate no such event with these failures), which Trend missed in their design and QA.
Since this is a consumer- or small-business oriented switch, it's unlikely that most customers will be pushing as much data through it as consistently as my client's environment, so perhaps that's a contributor. Otherwise, the only thing that stands out is Trend's "GREENnet" technology which reduces power utilization per port based primarily on cable length - perhaps there's something unusual in that technology that's having trouble either with my cables (all Cat6 500Mhz both for endpoints and for premise-punchdown with a maximum switch-to-client distance of 12 meters through two punchdowns, and short-run Cat6 500Mhz direct cabling to the server and other gear in the IDF closet itself), some other obscure item like MTU, a problem with heat (there is no dedicated AC in the IDF but the temperature remains constant at about 80F, which should be perfectly fine) or some combination of the above. Either way, given that everything in my architecture is designed correctly, was working perfectly with the SRW2008 24x7 for over a year, and, again, has worked perfectly with each of these Trend switches - until they die in this decidedly-odd fashion - I must conclude that there is a design flaw. While it's possible that I received a pair of bad switches (maybe there was an issue with one batch in manufacturing), TrendNet hasn't acknowledged such an issue. I'm certainly not going to waste any more time trying a third one of these.
For now, I've placed the SRW2008 back in service, chained to a dumb 10/100 switch for a few low-bandwidth endpoints, until I decide on a replacement device. Given Trend's overall-excellent reputation and reviews, I'm willing to give them another chance, but only with a device with a different architecture. The main contender right now is their TEG-160WS, which is managed, does NOT have GREENnet, and is only about $50 more. I'll write an update if I do go with that unit.
*2/13/12 Update: I ended up replacing the TEG-S16Dg with the latest "big brother" of the SRW2008, the Cisco Sg 200-18 18-PORT Gigabit. Despite the much higher price of admission, I wanted the management and monitoring capability (which is absurdly extensive given the target audience of small businesses) and the Cisco reliability to which I'm accustomed from my large datacenter implementations. For once, it seems like a manufacturer managed to integrate an acquisition - Linksys in this case - very well; while this product's origins were as a high-end Linksys and ended up as a low-end Cisco, the throughput, reliability (so far), configuration options, and other various bells and whistles make this switch a relative bargain. Note that it too offers the "green ethernet" option (I hadn't been aware of this initiative and how widespread it's become), but I've disabled it for now (+1 for highly-configurable managed switches); if the switch survives through the end of the month I'll turn it on to see what happens - I haven't yet discounted that there might be something unusual happening in the environment with an IP phone or desktop dumb switch that's causing the low-power / cable length detection to go wonky.
*6/26/12 Update: Except for configuration-change-necessitated reboots, the Cisco has been running 24x7 since I installed it with zero issues. So, unless Trend has finally acknowledged and/or fixed their design flaw, skip this and buy the Cisco.