Sanskrit distinguishes long vowels from short ones and the distinction is meaningful, so a Tibetan translator writing a Sanskrit name or a mantra could not simply ignore it. Tibetan has no vowel length of its own and its four signs make no provision for one.
The solution was to extend the signs rather than to invent new ones. A doubled vowel sign writes the long vowel: two greng bu for a long e, two na ro for a long o, and the same principle for i and u. For the long a, which has no sign of its own to double because a is the inherent vowel, the small a is subjoined beneath the letter instead.
This belongs to the same body of machinery as the reversed letters and the subjoined ha: a set of devices built beside the alphabet so that Sanskrit could be written exactly, without disturbing the thirty letters and four signs that a Tibetan child recites. All of it is taught after the alphabet and to a narrower group, and a reader of ordinary Tibetan may never need any of it.
The general shape of the solution is characteristic of the script. Where something new is needed, an existing element is modified, doubled, reversed or subjoined. Nothing is imported whole.
None of it changed the recitation. A Tibetan child still learns four vowel signs and thirty letters, and the long forms are taught later, with the rest of the transcription apparatus, to those who will have to read Sanskrit.