Genetically identical microbial cells typically display significant variability in every measurable property. In particular, highly abundant proteins - which can determine cellular behavior exhibit large variability in copy number among individuals. Their distribution has a universal shape common to different proteins and microorganisms; the same distribution shape is measured both in populations and in single-cell temporal traces. Moreover, different highly expressed proteins are statistically correlated with cell size and with cell-cycle time. These results indicate
coupling between measurable properties in the cell and buffering of their statistics from the microscopic scale. We propose a modeling framework in which the complex intracellular processes produce a phenotype composed of many effectively interacting components. These interactions, as well as the imperfect nature of cell division events, provide a simple model that reconstructs many properties of phenotypic variability on several timescales. These include fluctuating accumulation rates along consecutive cell-cycles, with correlations among the rates of phenotype components; universal and non-universal properties of distributions; correlations between cell-cycle time and different phenotype components; and temporally structured autocorrelation functions with long (∼ 10 generation) timescales.