The raised cosine pulse is essentially a windowed sinc function, designed with some roll-off in mind. Why do we use this particular window, as opposed to other windows such as Hann, Hamming, Kaiser, etc? All of these have zero ISI when applied to a sinc function, and have tunable frequency characteristics just like the raised cosine. Is the raised cosine optimal for digital comms in any sense?
There is no good reason, except that it's familiar and "good enough" for many applications. Many other better pulses have been proposed; see for instance "Nyquist filters in non-ISI transmission", by Ping-Kuen Lam, E.W. McCune and M.A. Soderstrand, DOI:10.1109/MWSCAS.1997.666198.