Résumé : Star-forming galaxies (SFGs) display a continuous distribution of specific star formation rates (sSFR) which can be approximated by the superposition of two log-normal distributions. The 1st of these encompasses the main sequence (MS) of SFGs, the 2nd one a rarer population of starbursts (SB). We show that the sSFR-distribution of SBs can be regarded as the result of a physical process (plausibly merging) taking the mathematical form of a log-normal boosting kernel and enhancing star formation activity. We explore the utility of splitting the star-forming population into MS and SB galaxies - an approach we term "2-Star Formation Mode" (2-SFM) framework - for understanding their molecular gas properties. Variations of star formation efficiency (SFE) and gas fractions among SFGs take a simple, redshift-independent form, once these quantities are normalized to the value of an average MS galaxy. The change in SFE for galaxies undergoing a starburst event scales supra-linearly with the SFR-increase, as expected for mergers. This implies a continuous distribution of galaxies in the Schmidt-Kennicutt plane that separates more clearly into loci for SBs and normal galaxies than observed in SFR vs. M* space. SBs with the largest deviations (>10-fold) from the MS, like many local ULIRGs, are not average SBs, but even rarer events with progenitors having larger gas fractions than typical MS galaxies. We statistically infer that the gas fractions of typical SBs decrease by a factor of 2 to 3 with respect to their direct MS progenitors, as expected to occur in short-term SFR-boosts during which internal gas reservoirs are drained more quickly than gas is accreted from the cosmic web. We predict variations of the CO-to-H2 conversion factor in the SFR-M* plane and provide evidence that the higher sSFR of distant galaxies is a direct consequence of larger gas fraction in these systems.