The present work reports results for: pd radiative capture observables measured at center-of-mass (c.m.) energies in the range 0--100 keV and at 2 MeV by the TUNL and Wisconsin groups, respectively; contributions to the Gerasimov-Drell-Hearn (GDH) integral in 3He from the two- up to the three-body breakup thresholds, compared to experimental determinations by the TUNL group in this threshold region; longitudinal, transverse, and interference response functions measured in inclusive polarized electron scattering off polarized 3He at excitation energies below the threshold for breakup into ppn, compared to unpolarized longitudinal and transverse data from the Saskatoon group. The calculations are based on a realistic Hamiltonian with two- and three-nucleon interactions and a realistic current operator, including one- and two-body components. The theoretical predictions obtained by including only one-body currents are in violent disagreement with data. These differences between theory and experiment are, to a large extent, removed when two-body currents are taken into account, although some rather large discrepancies remain in the c.m. energy range 0--100 keV, particularly for the pd differential cross section and tensor analyzing power at small angles, and contributions to the GDH integral. A rather detailed analysis indicates that these discrepancies have, in large part, a common origin, and can be traced back to an excess strength obtained in the theoretical calculation of the E1 reduced matrix element associated with the pd channel having L,S,J=1,1/2,3/2. It is suggested that this lack of E1 strength observed experimentally might have implications for the nuclear interaction at very low energies. Finally, the validity of the long-wavelength approximation for electric dipole transitions is discussed.