It matters greatly how we "observe". For instance, SDSS imaged the parts of the sky visible from their site in a range of bandwidths. Their filters are labeled u, g, r, i, and z. It is possible (and has been done quite well) to combine such imagery to simulate the appearance of broad-band visual images. Schmidt telescopes in the northern and southern hemispheres were used to make whole-sky surveys on photographic plates over a course of decades. The images were made in blue, red, and infrared bands. IRSA hosts those images, and supplies functionality to combine those images in multiple bands to simulate the appearance of broad-band images, though that is not as well-implemented as the SDSS method.
In each band, observations are subject to natural extinction (luminosity falls off as the square of our separation) and our perception of the observations is skewed by Malmquist bias, in which the brightest, most energetic objects at every redshift are over-represented while the fainter, more spread-out objects fail to show in our imagery.
Are there dwarf galaxies, low-surface brightness galaxies, or galaxies populated by old, red stars at high redshift? They may exist, but we would have great difficulty in detecting them because of natural extinction. The OP posted in philosophy, but there is plenty of hard science to apply to the question.