"John Larkin" ...
Do you think a right-angle trace is more likely to break than a curve
or bevel? I can't see why.

Some time before 1980 I did some PCB layout for the European IRAS satellite.
In those days the minimum inside curvature of a bend was specified (AFAIR
about 0.7 mm inside, mo problem since we taped the layout at 4x or 8x
scale): the reason given was better behaviour with thermal cycling. Outside
had to follow the inside with the track width, no sharp points allowed.

Arie de Muynck.