We critically review the basic assumptions of the standard model for the synthesis of the X-ray background (XRB) in the light of new data from ultradeep surveys by Chandra and XMM, resolving major parts of it. Important constraints come in particular from the observed redshift distributions of faint hard X-ray sources - showing large excesses at redshifts (z~ 0.8) much lower than expected by the synthesis models - and from their X-ray/optical/infrared spectral energy distributions combined with the infrared counts of type II AGNs. We find that hard X-rays and the mid-infrared appear to detect the same population of buried AGNs with peak emissivity around z~ 1. This analysis, although supporting the general scheme which interprets the XRB as due to absorbed AGNs with broad NH distributions, requires major revision of the other postulate of the XRB synthesis models: the AGN unification. We argue that the unification scheme based on a simple orientation effect fails at high redshifts, where galaxy activity is induced by strong interactions and mergers among gas-rich systems. This helps to explain the observational evidence that type I and II AGNs follow different evolutionary patterns, with type I quasars providing a very biased trace of this activity. Combined deep X-ray and infrared surveys consistently find that the Universe has experienced a violent phase of galaxy activity around z~= 1, probably related to the assembly of massive galaxies. This has involved both star formation (primarily sampled in the infrared) and obscured AGN fuelling (as detected in hard X-rays and mostly responsible for the XRB): our analysis implies that roughly 10 to 20 per cent of this activity has involved substantial AGN emission, this fraction probably reflecting the AGN/starburst duty cycle during the activation phase.