No geometric stress singularity then, good.

You have shown a single data point for the element size used for those results.  What is that element size?

Please reply with two plots: 1) maximum stress and 2) minimum life versus element size over a large number of sizes, at least 4 points

Use a constant ratio of element size.  A factor of 1.5 is good. Here is a series of element sizes in mm.

 1.00,  0.67,  0.44,  0.30,  0.20

If the points are not beginning to fall on a curve that looks like it will extrapolate smoothly to zero element size, keep going with smaller numbers.

You don't need the entire model to have this element size, only the faces adjacent to these max and min locations.