Large multidisciplinary research organisations face a well-documented tension: their breadth isprecisely what makes them suited to complex societal challenges, yet their disciplinary structure(centres, areas, evaluation panels) makes interdisciplinary work costly to initiate and hard to sustain.This paper presents a longitudinal institutional case study of the approach adopted by the SpanishNational Research Council (CSIC) — Spain’s largest public research organisation, with more than1,500 research groups in around 120 institutes — to address that tension without restructuring theinstitution. From 2018 onward CSIC developed a graded set of three collaboration instruments:Conexiones (light, bottom-up thematic networks), Interdisciplinary Thematic Platforms (PTI: mission-oriented, three-year, challenge-driven platforms with an ESFRI-derived design questionnaire) andExtended Platforms (PTI+: large-scale platforms with a four-line work structure, a consortium-agreement style governance and centralised financial management, created to absorb investmentsabove C10M). The design principles of each instrument, the formal texts that implemented them andthe governance rules are presented (dual coordination across centres, assembly/executive-committeemodel, two-thirds quorum, individual veto, VICYT ratification, Team Science and Open Sciencerecommendations). Publicly available documentary evidence on outcomes up to 31 August 2026is then compiled for the best-documented platforms and networks, including the PTI Salud Globalresponse to COVID-19, the QTEP quantum ecosystem, SusPlast pilot plants, the IPCC Interactive Atlasdelivered under PTI Clima, the Conexiones cohorts created in 2023 and 2025, and the 2026 migrationof the Teledetect remote-sensing platform into the EarthWatchHub Conexión. The documentary recordsupports four observations: (i) a pre-existing platform template enabled unusually rapid institutionalmobilisation during COVID-19; (ii) the platform architecture and its core governance principles wereadapted to an increase of roughly two orders of magnitude in financial scale; (iii) a graded portfolioallowed organisational migration between instruments instead of simple termination; and (iv) differentinstruments generated documented joint activities and outputs. The instruments were subsequentlyformalised, not discontinued, by the administration that succeeded the one that created them. A setof transferable design propositions is derived from the case, and the limitations of a documentaryanalysis without a counterfactual are discussed. The contribution is a longitudinal institutional caseshowing how a large disciplinary research organisation can superimpose a graded set of transversecollaboration structures on its existing organisational backbone without dismantling it.