When sigmaZ is negative, the integrand blows up around x = 0.3 and x = 2.7 and you have a singularity which you cannot integrate.
For example, just plot it:
sigmaZ= -0.1464*eye(4);
f = @(x)(arrayfun(@(x)(1./(det(sigmaZ./sin(x).^2 + eye(4)))),x))
ezplot(f)
When sigmaZ is positive, the integrand is always finite for x in [0 pi] and so you don't get any problem.
