Dear Jacek,
the sentence that you quoted by the paper of Rezzolla reads:
"... it is not easy to predict that there exist solutions behaving as waves."
which is different from:
"... it is impossible to predict that there exist solutions behaving as waves."
In fact, such solutions of exact equations have been found, see for example
H. Bondi, F.A.E. Pirani, and I. Robinson, Proc. R. Soc. London A 251, 519 (1959);
H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers, and E. Herlt, Exact solutions of Einstein's field equations, 2nd ed. (Cambridge Univ. Press, Cambridge, 2003) Secs. 24 and 31;
J. Ehlers and W. Kundt, Exact solutions of the gravitational field equations, in: "Gravitation: an introduction to current research", ed. by L. Witten (Wiley, New York, 1962);
Y. N. Obukhov, J.G. Pereira and G. F. Rubilar, Class. Quant. Grav. 26, 215014 (2009).
On the other hand, it is true that
"the concept of gravitational waves as solutions of (linearized) Einstein equations is valid only under some rather idealized assumptions such as: a vacuum and asymptotically flat spacetime and a linearized regime for the gravitational fields."
But, even if idealized, such assumptions can be realized with an excellent approximation. Take two neighbouring free-falling test masses within a vacuum cylinder in the Earth's gravitational field. For Einstein's equivalence principle they realize a vacuum and locally flat spacetime. This is EXACTLY the principle which is used for detect gravitational waves with interferometers. In fact, test masses suspended to pendulums represent an excellent approximation of free-falling test masses.
More, the linearized regime is an excellent approximation even for the gravitational field of the Solar System (historically Einstein used the linearized regime to test astrophysical predictions like the light deflection, the gravitational redshift and the precession of the Mercury's perihelion). The gravitational field of a gravitational wave is various order of magnitude lower than the gravitational field of the Solar System, thus, the linearized regime is an absolutely excellent approximation. In fact, high order terms are lower than 10^-42...
Best regards,
Ch.