A deliberate project
This was state policy before it was a tradition. In the early 9th century Tibetan decrees standardised how Buddhist vocabulary should be translated, and a Sanskrit to Tibetan glossary, the Mahāvyutpatti, fixed the terms so that different translators would agree.
Work was done in pairs or teams: an Indian scholar who knew the text and a Tibetan lotsāwa, a translator, who knew how to make it Tibetan. Colophons name both, along with the patron, which is why so much of this literature can be dated.
The second wave
After the collapse of royal support, the work resumed from western Tibet. Rinchen Zangpo, who lived from 958 to 1055, was sent to Kashmir by the king of Guge, came back with teachers and artists, and translated for decades. Atiśa arrived from the great monastery of Vikramaśīla in 1042, and for Tibetans his coming marks the restoration of the teaching.
So much was retranslated that Tibetans came to speak of new translations and old ones, and the schools themselves divide along that line.
What it saved
The translations were eventually organised into two collections: the Kangyur, the translated word of the Buddha, and the Tengyur, the translated commentaries. They were gathered and edited in the 14th century, Buton Rinchen Drub's work at Shalu being the most influential, and later printed from woodblocks at Narthang, Derge and elsewhere.
Because the translators worked so literally, following Sanskrit word order and using fixed equivalents, scholars can reconstruct lost Sanskrit works from the Tibetan with some confidence. A great deal of Indian Buddhist philosophy exists today only in Tibetan. The work is not finished: the 84000 project is translating the Kangyur into English, volume by volume, and publishes it free.
Why it matters
Tibetan Buddhist collections preserve a vast body of Indian Buddhist literature, including works whose Sanskrit originals were subsequently lost.
History and tradition
What historians can say
The Dunhuang manuscripts and the surviving glossary show that the standardisation was real administrative policy and not a later legend. The scale is documented in the catalogues: the Kangyur and Tengyur together run to over three hundred volumes.