In general operating systems do not deliberately shuffle processes to produce hard randomness; they're aiming for scheduling that lets all processes advance with some reasonable fairness. Libraries that consume this data for harvesting entropy are already at a disadvantage, relative to an operating system component for harvesting entropy.
Newer operating systems incorporate blocks of entropy derived from peripheral signals [e.g. mouse trajectories], although again, the distribution of typical data from any given peripheral is characteristic of that peripheral [and for a specific user, also a sort of cyber-biometric, in a manner similar to gait analysis], so you're still looking at biased samples.
In general operating systems do not deliberately shuffle processes to produce hard randomness; they're aiming for scheduling that lets all processes advance with some reasonable fairness. Libraries that consume this data for harvesting entropy are already at a disadvantage, relative to an operating system component for harvesting entropy.
Newer operating systems incorporate blocks of entropy derived from peripheral signals [e.g. mouse trajectories], although again, the distribution of typical data from any given peripheral is characteristic of that peripheral [and for a specific user, also a sort of cyber-biometric, in a manner similar to gait analysis], so you're still looking at biased samples.