The thermal limiter will throttle back the cpu speed from 1.5Ghz → 1.2Ghz → 1.0Ghz → 0.66Ghz as necessary to try to keep the temperature under 105C.
Normally under an extreme load it should stay below about 110C. At 114C it will log temperature warnings in dmesg, and at 118C it will shut down the system to prevent further overheating, and as sam mentions this will cause the LED to go red.
As this is a shutdown it requires a power cycle to boot again. The reason why it shuts down at 118C after all attempts to regulate the temperature have failed (cpu throttled back to 667 Mhz) is because the device will crash at about 122C, and this sometimes causes it to get extremely hot, while shutting down safely will let it cool right down, so this is a last line of defence to protect the device.
So if you are now seeing it shut down with a red led it is indeed overheating. What sort of environment is it exposed to ? I have mine inside a closed TV cabinet along with a TV, Stereo amplifier and Xbox one (inside cabinet about 26C) and mine doesn’t come anywhere near overheating even with cpuburn-a53 running. (A very stressful CPU load designed to heat up a cpu)
You could try running the following script via ssh:
watch -n 0.5 "echo -n 'temp: '; cat /sys/class/thermal/thermal_zone0/temp; echo -n 'cpu speed: '; cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq"
Leave this running, keeping an eye on it - it will show the SoC temperature (in millidegrees, so divide by 1000 for degrees C) and cpu speed, updating twice a second.
Above 105C the CPU speed will start to throttle back and it should not get any hotter than about 110C. If you see it continue to rise and hit 118C with the cpu speed at 667 and then the light goes red and the device shuts down it has definitely overheated, presumably due to an excessively hot environment or lack of ventilation.
A picture of where the device sits to provide some context may be helpful.