Equilateral
triangles. Dan Asimov asks how large a triangle will fit into a
square torus; equivalently, the densest packing of equilateral triangles
in the pattern of a square lattice.
There is only one parameter to optimize, the angle of the triangle to
the lattice vectors; my answer
is that the densest packing occurs when
this angle is 15 or 45 degrees, shown below.
(If the lattice doesn't have to be square, it is possible to get density
2/3; apparently this was long known, e.g. see Fáry,
Bull. Soc. Math. France 78 (1950) 152.)
Asimov also asks for the smallest triangle that will always cover at least
one point of the integer lattice, or equivalently a triangle
such that no matter at what angle you place copies of it on an integer lattice,
they always cover the plane; my guess is that the worst angle is parallel
and 30 degrees to the lattice, giving a triangle with 2-unit sides
and contradicting an earlier answer to Asimov's question.