Since the two lights are so different, let's discuss them separately. First the Zebralight. I'll discuss the Acebeam in another post.
You didn't say which model you had, but they only one they offer that claims XHP50.2 is the SC600w Mk IV.
If the batteries are the problem, then you would notice that the light dims steadily as the batteries discharge. This would be most noticeable right off charge.
Another possibility is that the LEDs are overheating, or are just garbage LEDs. If they were 'recent' puchases I'd assume the runtime has been on the order of hours or tens of hours. The dies would have to be pretty badly overheating to degrade that rapidly. Do you see any discoloration on the dies?
The XHP50.2 is rated at a max drive of 3A in the 6V configuration, or 1.5A in the 12V config. Either way that's 18W. To get 2300 lm, you'd have to be pushing 18W and getting 128 lm/W, which the XHP50.2 can't quite do, especially at 80 CRI and 4500K. So to get anything like the claimed 2300 lumens even at the emitter (OTF would be considerably less), you'd have to push the chip quite hard. In a light this size, this is not sustainable, as the light would quickly become too hot to hold.
They claim PID temperature control, which would reduce the brightness (and thus the heating) as the temperature reaches the maximum. Thus you would expect, if running it on high, that it would come on bright, stay at an even brightness (at least to the eye) for a short time, then rapidly dim quite noticably. I would expect it to take several minutes, perhaps 5 or 10, to dim to a relatively steady level. The body would over the same time warm to a relatively steady temperature.
Zebralight picked an operating temperature, and claim to have a controller that can maintain it, and give you +/- 5°C control over it. But they don't tell you what that temperature is, or where or how it's measured (hint: it's NOT measured at the die). These are important, as high design temperature would lead to short life, and poor design or assembly could lead to significantly higher than expected die temperatures, and thus significantly less than design life. I also notice they don't publish a design life.
It just occurred to me, another thing to consider is that if the power level (and thus the light output) is actually limited by die temperature, then then temperature of the environment would have an impact on brightness. The cooler the environment, the more power you can dissipate while maintaining a steady temperature. So maybe the difference you are seeing is seasonal.