The galaxies missed by Hubble and ALMA: the contribution of extremely red galaxies to the cosmic census at 3<z<8

Using deep JWST imaging from JADES, JEMS and SMILES, we characterize optically-faint and extremely red galaxies at $z>3$ that were previously missing from galaxy census estimates. The data indicate the existence of abundant, dusty and post-starburst-like galaxies down to $10^8$M$_\odot$, below the sensitivity limit of Spitzer and ALMA. Modeling the NIRCam and HST photometry of these red sources can result in extreme, high values for both stellar mass and star formation rate (SFR); however, including 7 MIRI filters out to 21$\mu$m results in decreased mass (median 0.6 dex for log$_{10}$M$^*$/M$_{\odot}>$10), and SFR (median 10$\times$ for SFR$>$100 M$_{\odot}$/yr). At $z>6$, our sample includes a high fraction of little red dots (LRDs; NIRCam-selected dust-reddened AGN candidates). We significantly measure older stellar populations in the LRDs out to rest-frame 3$\mu$m (the stellar bump) and rule out a dominant contribution from hot dust emission, a signature of AGN contamination to stellar population measurements. This allows us to measure their contribution to the cosmic census at $z>3$, below the typical detection limits of ALMA ($L_{\rm IR}<10^{12}L_\odot$). We find that these sources, which are overwhelmingly missed by HST and ALMA, could effectively double the obscured fraction of the star formation rate density at $4<z<6$ compared to some estimates, showing that prior to JWST, the obscured contribution from fainter sources could be underestimated. Finally, we identify five sources with evidence for Balmer breaks and high stellar masses at $5.5<z<7.7$. While spectroscopy is required to determine their nature, we discuss possible measurement systematics to explore with future data.