The field of blazar γ-ray astrophysics is living a quite exciting time. AGILE and Fermi-LAT provide an unprecedented dynamical view of these sources over several timescales. Cherenkov arrays on ground are revealing an unexpected variety of phenomena, such as the ultra-rapid flares and the very-high energy emission of FSRQ. The current observational evidence is probably just the top of the iceberg.
In particular, the current sparse observations of ultra-rapid variability events in the VHE band leaves several questions unanswered. In particular the duty cycle (i.e. the frequency) of these events is a key information, potentially able to test and to constrain some of the proposed scenarios. Of course, a proper evaluation of the duty cycle would require a dense and frequent monitoring of the most promising targets. The events detected so far happens during high-state, flaring states of the sources. It would be quite interesting to know whether this kind of phenomena can also occur in states of low/quiescent activity. Particularly relevant for our understanding would be the detection of such events in other bands, such as optical or X-ray. So far, observations have not revealed a phenomenology as that observed in the high-energy band. To check whether ultra-rapid variability appears also in these bands, strictly simultaneous observations in different bands are required.
The detection of very-high energy photons from FSRQ is a quite informative result already changing our view of these sources. We know that, at least occasionally, the emission region of jets in FSRQ can leave the BLR. We do not have a clear explanation of the mechanisms triggering these states and how often they occur. The extension of the detection within the TeV band could provide other important information, especially if a cut-off induced by absorption on the IR radiation field is revealed. In principle such observations would even provide us the opportunity to probe the IR emitting regions of the AGN.
Although in this paper I focused on the two issues discussed above, it is important to mention two other topics potentially very important related to the high-energy emission from blazars. The first topic is that involving the so-called extreme BL Lacs, characterized by several peculiar aspects, such as an extreme energy of the synchrotron peak maximum (in some cases exceeding 10 keV), an exceptionally hard GeV-TeV continuum, locating the IC peak above several TeV (to be compared with a typical IC peak energy of 100 GeV) a limited variability at high-energy. An interesting scenario interprets these peculiarities invoking the existence of a beam of ultra-high energy hadrons escaping from the jet and producing the γ-ray emission on the way toward the Earth through Bethe-Heitler and photo-meson cooling . A robust and testable prediction of this model is the presence of a detectable hard γ-ray tail extending above 10 TeV  even for sources located at redshift close to 1 . In the standard models the EBL absorption would prevent the detection of photons with these extreme energies. A detection of even few photons with E > 10 TeV by CTA would therefore be a spectacular confirmation of the hadronic beam model.
Finally I would like to mention the potential role of blazars as cosmological tracers of the evolution of the most supermassive black holes. As stressed by  the most powerful FSRQ are indeed hosted in systems with a mass of the black hole exceeding 109 solar masses. The detection of these FSRQ at high redshift (beyond 2-3) is thus a powerful tool to make a census and follow the evolution of massive black holes . Powerful FSRQ display an IC peak with a maximum in the soft gamma-ray band, around few MeV. Indeed they usually escape detection by Fermi-LAT (characterized by a low energy limit of 100 MeV) but appear in hard X-ray surveys such as that performed by BAT onboard Swift. A powerful detector of MeV blazars such as the proposed e-ASTROGAM woulf be the ideal instrument for this kind of studies.
I thank the organizers for the invitation. It is a pleasure to thank G. Ghisellini and G. Bonnoli for years of fruitful collaboration. I am grateful to L. Sironi for stimulating discussions. This work has been partly founded by a PRIN-INAF 2014 grant.