Long i is a grammatical letter more than a lexical one. It marks a large class of feminine stems, it stands in the endings that build causatives and denominatives, and it appears in a set of very common inflections, so the sign is met constantly while the number of roots that actually contain the vowel is modest. The consequence for a scribe is that most occurrences of the mark are predictable from the shape of the word rather than from memory, and most errors in it are grammatical errors rather than slips of the hand.
Its dependent sign mirrors the short one closely enough that the two are separated by a small difference of stroke. In a careful hand the distinction is plain; in a rapid one it is not, and this is one of the pairs a reader of manuscripts learns to judge from context. Sanskrit uses vowel length to distinguish words that are otherwise identical, so the difference is not a nicety. Getting it wrong produces a different word, and in a doctrinal text a different claim.
The independent letter, like every vowel in the script, is written only where a word begins on the sound, which happens more rarely for this vowel than for most. That rarity is a fact about the language rather than about the script, and it is why the independent forms of several vowels are found chiefly in the letter tables and hardly at all in running text. The table is not a frequency list and was never meant to be read as one.