When does long exposure noise reduction help on Canon cameras?

The real variable is not exposure time. It is the total amount of light that is allowed to enter the camera during the exposure time used.

If the scene is bright enough to give a medium to bright exposure at 30 seconds, then the need for LENR will be less than if the scene is very dark and most of the frame is still fairly dark after a 30 second exposure. This is because our perception of noise is not based upon the amount of noise in an image. It is based on the ratio of the amount of signal (light) to the amount of noise in the image. Brighter exposures, at whatever exposure times, mask the noise that is always present in a digital image better than darker exposures do.

That's why I find "noise tests" where the lens cap is left on the camera to be more or less useless. Without any light there is no signal to compare to the noise. If a camera has larger pixels with higher full well capacity, it can conceivably demonstrate more "noise" during a "lens cap" test and still have a better SNR when a properly exposed image that exploits the full dynamic range of the camera is used versus another camera that generates less read noise but has smaller pixels, lower full well capacity/dynamic range, and thus a lower SNR.

A secondary variable is the amount of heat that builds up in the sensor during long periods when the sensor is energized. The sensor could be energized due to long exposures, repeated shorter exposures for a long period of time, or use of Live View for composing still images or for shooting video. Because a sensor will heat up during a longer exposure there will be more actual noise in the image. If the amount of signal is the same, the signal-to-noise ratio (SNR) will be lower and the image will look noisier than if the sensor were cooler. But the real culprit is still the ratio of signal (light) to noise, not the absolute value of the amount of noise generated by the camera.