It's most likely for round-trip compatibility with another encoding. There are many Unicode codepoints that simply represent combinations of other codepoints. If you don't care about round-tripping, just normalize everything to NFKC or NFKD (the difference being that accented letters like á are one codepoint in NFKC and two codepoints for the base letter and combining mark in NFKD).

For lossless round-tripping with legacy character sets. This is one explicit design goal of Unicode.

So we can make search that tiny bit more complicated.