Siddham was written by hand for the whole of its working life, in India, along the routes east, in Chinese monasteries and in Japanese ones, and a script transmitted that way accumulates variation as a matter of course. The modern encoding of Siddham records the fact directly: alongside the standard shape of several letters it carries alternate forms, present because they occur in the manuscript record often enough that a text cannot be represented without them. Those alternates are ways of drawing the same letter and not additional letters, and the distinction matters to anyone trying to say how large the syllabary is.
Variation of that kind carries information. A hand that keeps an older shape after the Indian tradition has moved on tells you something about when and from where its exemplar came, and East Asian copies are full of such retentions, because the transmission ran by exemplar rather than by continuous contact. A Japanese scribe of a much later century might be reproducing faithfully a form that had gone out of use in India long before he was born.
It also creates the standing risk of any copied tradition. Two letters whose signs differ by a single small stroke are the ones a tired copyist merges, and the pairs at risk in Siddham are the short and long vowels, whose marks differ minimally, and the letters for the voiced labial stop and the labial semivowel, which are drawn alike in many hands. Where a text has passed through several generations of copying, an edition of it is partly a reconstruction of which of those a scribe intended.