What's wrong with the Covariance Matrix of my Training Data? Is my data suitable for use by Machine-Learning algorithms?

Question

I have this training dataset, made from simulated values of sensor readings. The database contains 13960 samples, with an equal no of samples belonging to 4 classes. The sam