If I understand your question correctly, your new array is actually a 2-dimensional array, with the first dimension representing the time a sample was taken, and the second dimension representing the samples of a sinusoidal.
To process your new array via the FFT, simply ignore the time dimension, and use only the samples dimension (the sinusoidal data samples).
The Nyquist rate still applies in this case. Your sinusoidal must be sampled at twice (or more) the maximum frequency that you hope to resolve via the FFT.
The FFT uses the sampling frequency to properly scale the frequency axis of the FFT output.
You can learn all about the Nyquist–Shannon sampling theorem here: Sampling theorem