We present photometric and spectroscopic follow-up of a sample of extragalactic novae discovered by the Palomar 60-inch telescope during a search for "Fast Transients In Nearest Galaxies" (P60-FasTING). Designed as a fast cadence (1-day) and deep (g < 21 mag) survey, P60-FasTING was particularly sensitive to short-lived and faint optical transients. The P60-FasTING nova sample includes 10 novae in M31, 6 in M81, 3 in M82, 1 in NGC2403 and 1 in NGC891. This significantly expands the known sample of extragalactic novae beyond the Local Group, including the first discoveries in a starburst environment. Surprisingly, our photometry shows that this sample is quite inconsistent with the canonical Maximum Magnitude Rate of Decline (MMRD) relation for classical novae. Furthermore, the spectra of the P60-FasTING sample are indistinguishable from classical novae. We suggest that we have uncovered a sub-class of faint and fast classical novae in a new phase space in luminosity-timescale of optical transients. Thus, novae span two orders of magnitude in both luminosity and time. Perhaps, the MMRD, which is characterized only by the white dwarf mass, was an over-simplification. Nova physics appears to be characterized by quite a rich four-dimensional parameter space in white dwarf mass, temperature, composition and accretion rate.

The distribution of pairwise, relative peculiar velocities, $f(u;r)$, on small nonlinear scales, $r$, is derived from the Press--Schechter approach. This derivation assumes that Press--Schechter clumps are virialized and isothermal. The virialized assumption requires that the circular velocity, $V_c \propto M^{1/3}$, where $M$ denotes the mass of the clump. The isothermal assumption means that the circular velocity is independent of radius. Further, it is assumed that the velocity distribution within a clump is Maxwellian, that the pairwise relative velocity distribution is isotropic, and that on nonlinear scales clump-clump motions are unimportant when calculating the distribution of velocity differences. Comparison with $N$-body simulations shows that, on small nonlinear scales, all these assumptions are accurate. For most power spectra of interest, the resulting line of sight, pairwise, relative velocity distribution, $f(u_{\rm r})$, is well approximated by an exponential, rather than a Gaussian distribution. This simple Press--Schechter model is also able to provide a natural explanation for the observed, non-Maxwellian shape of $f(v)$, the distribution of peculiar velocities.