We study the prospects of observing the non-resonant di-Higgs pair productionin the Standard Model (SM) at the high luminosity run of the 14 TeV LHC(HL-LHC), upon combining multiple final states chosen on the basis of theiryield and cleanliness. In particular, we consider the $b\bar{b}\gamma \gamma,b\bar{b} \tau^+ \tau^-, b\bar{b} WW^*, WW^*\gamma \gamma$ and $4W$ channelsmostly focusing on final states with photons and/or leptons and study 11 finalstates. We employ multivariate analyses to optimise the discrimination betweensignal and backgrounds and find it performing better than simple cut-basedanalyses. The various differential distributions for the Higgs pair productionhave non-trivial dependencies on the Higgs self-coupling ($\lambda_{hhh}$). Wethus explore the implications of varying $\lambda_{hhh}$ for the most sensitivesearch channel for the double Higgs production, \textit{viz.},$b\bar{b}\gamma\gamma$. The number of signal events originating from SMdi-Higgs production in each final state is small and for this reasonmeasurement of differential distributions may not be possible. Furthermore, weconsider various physics beyond the standard model scenarios to quantify theeffects of contamination while trying to measure the SM di-Higgs signals indetail. In particular, we study generic resonant heavy Higgs decays to a pairof SM-like Higgs bosons or to a pair of top quarks, heavy pseudoscalar decayingto an SM-like Higgs and a $Z$-boson, charged Higgs production in associationwith a top and a bottom quark and also various well-motivated supersymmetricchannels. We set limits on the cross-sections for the aforementioned newphysics scenarios, above which these can be seen as excesses over the SMbackground and affect the measurement of Higgs quartic coupling. We alsodiscuss the correlations among various channels which can be useful to identifythe new physics model.