The problem is that so many LED companies lack real quality/test discipline and experience. I drove north of Chicago and counted 5% off all freeway LED standards dead. Every day that I spent at least 30 minutes driving in Toronto, (which is very easy to do) I see at least 1 different new car, again every day, with a dead LED headlamp or perimeter string. This is for all brands including Audi, the pioneers of LEDs in cars.
Why? I can tell you in a nutshell. Lack of Test Engineering / Quality Management experience. Yet the auto industry has is second best in this regard, compared to Aerospace.
When I was in the HDD TEst Engineering business, I quickly realized these were the most complex instruments on earth, but not as harsh environments as Nuclear reactors or Space environment. Yet they have thousands of critical tolerances; contamination avoidance, thermo-magnetic dependancies, vibe&shock reduction methods for tracking, many anti-resonant controlled servo loops, critical HDA aerodynamics with heads flying at 100kph on an xx nanometer air bearing surface with lateral seek times of milliseconds to a position with x micron position error, adjacent track interference, susceptible to right angle rotational impacts, nanoparticle media contamination, high humidity head stiction, etc etc etc.
It takes an extreme set of skills and experiences to perform these tasks. All it takes is one critical review of the DVT and some boss to say we can't afford to do ORT or someone to overlook supplier quality tests and all of a sudden, a simple cost reduction or material change results in process "escapes" .
I can give you the benefit of my experience doing Design Validation Testing (DVT) and ORT for 5 years in a row and off and on for 30 years. Very few companies have this experience or perform this level of diligence necessary to prevent escapes.
Most LED suppliers allow Tj of the junction to rise to 125'C in the maximum ambient. The ambient is not the normal temeprature range of -40'C to +40'C, but the ambient of the heatsink under the sun with 1200 W/ sq.m. of solar power and no wind at max humidity.
Dysan has done a proper Engineering design from my quick review, to improve the area for passive cooling with good materials (sintered copper). Bravo, but only 85 Lumens per watt?. same as my much cheaper better Philips true daylight tri-phosphor 5000k tubes
This Dysan design may result in their heatsink temp rise being 15-20 deg lower than the standard adopted by many companies for Tj rise. From Arhennius Law we know a 10'C drop in Tj rise above ambient will reduce MTBF by x2. Thus I estimate Dysan's design of Tj rise is 15'C cooler than the worst case designs which spec 50kh and they would have to validate that number using large qty of units and measure Lumen deg. rates projected to LM70.
But what we dont know is how they do ORT and how do they know how much margin exists in shearing the Anode connection with the ultrasonically cold-welded micron size gold wire. THIS IS THE MOST COMMON failure mode in LEDs. It can occur for many reasons. Some designs use dual wire bond. But all must survive the extreme thermal stress from different coefficients of material expansion not get "wounded" from shock, vibration, ESD, thermal shock, transient overpower etc etc.
I do know I could buy 4x tri-phosphor 4' FL 32 or 28W tubes for at least half the price and twice the Lumen output, with the same efficacy in the mid 80LPW but slight better CRI by 2-5% as LEDs do not have triPhosphor bands, only 1 blue & 1 phosphor gets CRI=80-82 depending on CCT. These will last 50kh if left on 24x7 but only 30kh if cycled 3 times a day.
all for now, must fix son-inlaw's 92Camry with stuck thermostat from excess leakstop that I added to rad. Then hopefully he can get a job.