This is idea propose an alternative from classical means (like observing nuclear fading*) to get random sequences and generate random numbers : observing and using exploding Supernovaes as a way to get timed and spatialized random sequences to generate random numbers and sequences.

Why ? Because you get to observe the Supernovae at the time of its explosion — if you did not spot it earlier with your observatories (through star's light pre-supernovae deformations).

This means that at the technical limits of your horizon of observation — or the limit of the observable universe visible by your (radio)telescope — you will observe appearing dots of light right at the relative moment of the Supernovae explosion, that could not be observed before due to the design, technical or material limitations of the (radio)telescope.

The advantage of this technical is that by increasing the telescope's observation horizon (or its technical limitations — its observable universe), you'll reach relatively deeper in space, meaning that new unprecedentedly seen Supernovaes will be then observable.

Such as radioactive decrease, there might be an exponentially visible effect about the amount of observable Supernovaes as you reach relatively deeper in space. This would increase the statistical chances of observing dying stars (Supernovaes) — an exponential increase — thus recreating a coherence in the randomness numbers and sequences obtained by this mean.

*  See Schrödinger's cat experiment from 1935 according to Wikipedia. Quote : 

Schrödinger’s Cat is a thought experiment devised by the Austrian physicist Erwin Schrödinger, which he designed to illustrate a paradox of quantum superposition wherein a hypothetical cat may be considered both alive and dead simultaneously because its fate is linked to a random event that may (or may not) occur (sic : radioactive decay).