Why the Madagascar question recurs (five real threads, none load-bearing)

Readers keep reaching for Madagascar even after the Karana-are-Gujarati exclusion is stated. The pull is not irrational; five real threads keep it alive, and none are load-bearing enough to make Madagascar an Ecumene node. (1) INDIAN OCEAN TRADE MEMORY — Chola and later Tamil mercantile networks moved across the western Indian Ocean rim (East Africa, Comoros, Mascarenes), and Madagascar sits inside that rim geographically. (2) AUSTRONESIAN SUBSTRATE CONFUSION — Malagasy is an Austronesian language from the Barito family (Borneo origin); pseudoscience occasionally folds this into 'lost Tamil / Kumari Kandam' framings, which TLTE explicitly refuses (see tlte-cite:ecumene-connolly-2023 and the naming lock that the organ is 'Ecumene', never 'Kumari Kandam'). (3) RÉUNION / MAURITIUS ADJACENCY — the two confirmed Tamil indenture nodes nearest to Madagascar are Réunion and Mauritius; estate-labour and Hindu festival circuits did touch Madagascar's east coast in small numbers via those islands, but never produced a settled Tamil community. (4) KARANA ≠ TAMIL but are South Asian Muslim — to a casual reader 'South Asian Muslim trading community in the Indian Ocean' can sound like it might overlap with Tamil Muslims of the Eastern Province; it does not (different language: Gujarati not Tamil; different origin port: Kutch/Surat not Kayalpattinam; different denomination cluster). (5) SMALL CONTEMPORARY HINDU TAMIL PROFESSIONAL PRESENCE IN ANTANANARIVO — real but uncited at scale; not an Ecumene node by any standard the rest of the Atlas uses. TLTE will add Madagascar as a node ONLY if Tier-A sources later document Tamil-specific custodianship there. Until then the exclusion stands, and this recurrence note exists so the question stops looping. Lives at /ecumene/methodology/madagascar §'Why this question recurs'.