The freedom to vary $\lambda$ is severely constrained: For a nontrivial solution of the Killing spinor equation you need either a real and constant $\lambda$ or a function $\lambda$ with purely imaginary values [Lichnerowicz (1987)].
The generalization of the Killing spinor equation to nonconstant imaginary Killing function $\lambda$ has been worked outstudied in
H.-B. Rademacher, Generalized Killing spinors with imaginary Killing function and conformal Killing fields, Lecture Notes in Math. 1481 (Springer, Berlin, 1991).
A nonconstant $\lambda$ is only possible if its real part is identically zero.
A more recent paper along these lines is Complex Generalized Killing Spinors on Riemannian Spin manifoldsSee also (2013)Eigenvalues of the Dirac operator, Twistors and Killing Spinors on Riemannian Manifolds.