With the rapid deployment of hybrid mega-constellations, efficient resource allocation (RA) in multi-satellite collaborative (MSC) networks has become critical to support diverse, delay-sensitive, and computation-intensive services. This paper presents a structured three-tier collaboration framework, intra-orbit, inter-orbit, and heterogeneous satellite collaboration to transform static RA into adaptive, multi-timescale optimiza-tion. We analyze three fundamental challenges that distinguish MSC networks from terrestrial systems: highly dynamic topology, multi-dimensional onboard resource coupling (power, spectrum, computation, and caching), and complex three-dimensional (3D) interference. For each challenge we survey optimization tech-niques (convex programming, game theory, and deep reinforce-ment learning) and key enabling mechanisms including dynamic spectrum management, joint power/beamforming control, rout-ing and handover, computing resource scheduling, RIS-assisted beamforming, and advanced multiple access. Furthermore, a case study of RA under the proposed MSC framework is presented and realistic constraints are discussed. Finally, we highlight open problems, collaboration formation, green communications, security/privacy, and outline directions for future research.