[MatthiasWandel]

I haven't tested it explicitly, butI consider the magick mortise and tenon (or dowel) ratio to be around 4:1 or 5:1. A tenon longer than that will break off (wood failure), shorter than that, and it will pull out (joint failure). I suppose grooves on the dowel would help the glue a bit, but they also weaken the dowel.

[logfromblammo]

But your failure analysis showed that the point of failure was not the dowel, or the glue, but the wood near the wood-glue interface. The wood failed at planes parallel to the glue plane, such that the glue was left holding on to little splinters. If the glue did not form a plane, whenever the wood cracked parallel to the glue, those cracks would be misaligned, such that they would not immediately lead to joint failure by forming a shear plane. But it still has to be close enough to a plane that you can make the join in the first place. Grooving or dimpling one of the surfaces would be like nailing the dowel to the hole from the inside of the dowel as the glue dries, or adding o-rings that prevent lateral movement to a piece that requires lateral movement for assembly. The glue spikes/rings would have to shear off, or the wood would have to crack from the outermost extent of the spikes/rings.

Weakening the dowel is okay, since it wasn't the dowel that failed. It's probably more important to rough up the inside of the hole anyway, since the failure photos mostly showed that the glue was still stuck to the dowels, and it was wood from the inside surface of the hole that broke away. And the end-grain in the rail somehow frustrated the cracks better than the grain in the post.