Due to increase in forward bias voltage, the intensity of light increases but after a particular value the intensity decreases. Why?
The light output of a LED is pretty linear with the current through it, over its normal operating range. Light does usually drop off from linear with current at the high end. Sometimes that high end is not included in the normal operating range, so the graph you see in the datasheet will be linear. Common T1-3/4 20 mA indicator LEDs are usually linear over their 0-20 mA operating range. LEDs for lighting are pushed harder and operated closer to the limits, so you usually do see a smaller increase in light for the same current increase at high currents.
Note the D in "LED". These things are diodes. That means voltage will change little as a function of current. However, that also means the current can change drastically as a function of voltage. Furthermore, there is enough variation in forward voltage from part to part and due to temperature that you can't pick any one voltage to run a LED at. Basically, LEDs should only be driven with controlled current, letting the voltage come out to whatever it comes out to. Don't drive a LED with a fixed voltage. You are asking for thermal runaway, and there is no way to pick the correct voltage up front to get anywhere near the maximum light outut.
Therefore, asking why strange things happen when you force a particular voltage accross a LED is pointless. It is very easy to operate it outside its specified range by doing that, so there is no longer any guarantee what the device might to. Most likely, you overstressed the LED to the point that it had too much current thru it. That makes the die very hot, and can cause decreased light output and often cause a noticable color shift too. At that point the LED is on the edge of total failure, and has probably been irrevesably damaged. If you go back and measure its efficiency at intended currents, it will probably be lower than what it was originally.
Well LED internal quantum efficiency is highly temperature dependent, and generally even at CONSTANT CURRENT drive, the light will fall off at a higher junction Temperature.
Older technology, gallium arsenide phosphide (GaAsP) red LEDs generally are fairly linear with current (light output), and usually more efficient at higher currents, AT A CONSTANT JUNCTION TEMPERATURE. This occurred, because there were significant non-radiative recombination currents, which lowered efficiency at lower current densities. So those LEDs benefit from high current short duty cycle pulsing, which keeps the junction temperature down, but operates the LED at its higher efficiency higher current condition.
Modern gallium nitride LEDs, don't exhibit the same behavior and generally don't work better at higher current densities, and lower duty cycles.
So if you increase their current, and let them get hot, the light will go down. GaN material is a very high temperature semi-conductor, so the higher junction temperature operation won't necessarily kill the diode.
The older LEDs would die under this condition, because the 1 mil gold bond wire (or aluminum) will melt. Gan LEDs usually aren't wire bonded, so don't have that problem.