The facts of sandhi are treated comprehensively in all of the grammars, such as Macdonell (1916:20-39) and Whitney (1889:34-81), so I shall merely summarize the facts in this table.
The most striking fact about consonant sandhi is that with very few exceptions, it is a word-final consonant that is affected, not a word-initial one. In contrast to all the changes listed in the table for the final consonant, the only thing that happens to onsets is an occasional timing phenomenon: a stop may be inserted between nasal and sibilant, or an h after a stop may become the homorganic (but aspirated) stop, or an s" after a t (including excrescent t after n) may become ch. These are all caused by a slight delay in releasing the oral closure after a stop, and were all considered optional changes by the Prâtis"âkhyas. As Cho (1990) notes, this broad pattern of neutralization and assimilation is well documented for the ends of words, and indeed for the ends of syllables in general.