In cosmology, baryon acoustic oscillations (BAO) are regular, periodic fluctuations in the density of the visible baryonicmatter (normal matter) of the universe. In the same way that supernova provide a "standard candle" for astronomical observations,[1] BAO matter clustering provides a "standard ruler" for length scale in cosmology.[2] The length of this standard ruler (~490 million light years in today's universe[3]) can be measured by looking at the large scale structure of matter usingastronomical surveys.[3] BAO measurements help cosmologists understand more about the nature of dark energy (which causes the apparent slight acceleration of the expansion of the universe) by constraining cosmological parameters.[2].

For a large part of cosmic history, the a(t) curve of the standard model is very near to linear, so it is not too surprising that a large part of the data fits a linear expansion curve fairly closely. The problem is: which other physical theory predicts that?

Seems to me that my criticisms from that thread stand.

The thing is, professional cosmologists generally don't use a single type of data when determining which model better fits the data. Pretty much every talk that I've ever gone to that describes the evidence for dark energy makes use of a plot similar to this one: http://supernova.lbl.gov/union/figures/Union2.1_Om-Ol_slide.pdf

This plot is a bit of an old one, and there is even more data available now. What this is showing are the error contours from three different types of data, with the matter density fraction on the horizontal axis and the cosmological constant density fraction on the vertical axis. Note that each individual piece of data doesn't actually constrain dark energy all that well: the tight constraints come from combining them all together. The three data types are the CMB, BAO, and supernovae. The CMB's primary constraint relevant to this particular plot is on the spatial geometry: it says that the universe is very nearly flat. The Baryon Acoustic Oscillation (BAO) data, by contrast, mostly just constrains the matter density fraction. The supernova data constrain the ratio of matter density to dark energy density, but provides almost zero constraint on the curvature.

Taken together, these three data sets converge on the same location in the plot. That's the key point, and is why we can be pretty sure that dark energy actually exists. There are potentially two ways out of this at the current time:
1. There's a large, unaccounted-for systematic error that makes it so that these different data sets all converge tightly to the same location in parameter space, but converge to the wrong location. Some have suggested that the fact that most of these calculations usually assume the universe is homogeneous and isotropic, when it definitely is neither, might have something to do with this. But these alternative explanations have so far all failed in the face of more data.
2. There's some other model of gravity that explains why it looks like there's a cosmological constant when there actually isn't one. I don't think there's any coherent alternative to General Relativity that has been proposed that actually works here.

These two options only really still exist because there remains the possibility that there's something we haven't thought of.

This paper, by contrast, just says that, "Hey, when we throw out most of the data, the case for an accelerated expansion becomes rather weak!" Well, of course it does.

If I take a Milne model and add baryons at 3% of critical density ([itex]\Omega_b=0.03[/itex], no dark matter and no [itex]\Lambda[/itex]), Friedmann's equation gives the following "Distance-redshift" curve for comparison with the standard (LCDM) parameters. Incidentally, it gives a Standard age of ~13.7 Gyr.

One can see that this supports Chalnoth's comments of gross deviations at very early times.