One of the driving problems in the CSP area is the Dichotomy Conjecture, formulated in 1993 by Feder and Vardi [STOC'93], stating that for any fixed relational structure G the Constraint Satisfaction Problem CSP(G) is either NP--complete or polynomial time solvable. A large amount of research has gone into checking various specific cases of this conjecture. One such variant which attracted a lot of attention in the recent years is the LIST MATRIX PARTITION problem. In 2004 Cameron et al. [SODA'04] classified almost all LIST MATRIX PARTITION variants for matrices of size at most four. The only case which resisted the classification became known as the STUBBORN PROBLEM. In this paper we show a result which enables us to finish the classification - thus solving a problem which resisted attacks for the last six years. Our approach is based on a combinatorial problem known to be at least as hard as the STUBBORN PROBLEM - the 3-COMPATIBLE COLOURING problem. In this problem we are given a complete graph with each edge assigned one of 3 possible colours and we want to assign one of those 3 colours to each vertex in such a way that no edge has the same colour as both of its endpoints. The tractability of the 3-COMPATIBLE COLOURING problem has been open for several years and the best known algorithm prior to this paper is due to Feder et al. [SODA'05] - a quasipolynomial algorithm with a n^O(log n / log log n) time complexity. In this paper we present a polynomial-time algorithm for the 3-COMPATIBLE COLOURING problem and consequently we prove a dichotomy for the k-COMPATIBLE COLOURING problem.