I have an application for a threaded rod separating two plates that are sheared relative to one another.  The image below shows a test jig in which the threaded rod is simulated by a smooth rod "bonded" (as opposed actually threaded) into equal-diameter holes in the two blocks:

Here's a blow-up (for a particular mesh-control density) of the bottom joint, where the maximum stress exceeds the 20,000 psi yield strength of the steel:

I observe that the high-stress coverage decreases in area and increases in magnitude, apparently without limit, as mesh density is increased near the right-angle corner.  I'm not trained as an engineer, but I understand that this behavior is typical of "re-entrant edges," where the theoretical stress can be unbounded.  I have also modeled this configuration with fillets and observed that the stress converges nicely.  Unfortunately my application does not permit a fillet or other stress-relieving feature around the actual joints.

So my questions are:

1) What is the most realistic way to simulated such a threaded joint?

2) How can one establish a practical failure criterion in order to choose bolt diameter, etc? -- John Willett