Why is a space a %20 in one place and a + in another?
Both are the same space. A query string written as application/x-www-form-urlencoded uses + because a form post once sent spaces as plus signs, and a URL path or a fragment uses %20 because + there is only a plus. encodeURIComponent always writes %20, so a value that goes into a query string in a browser is safe with either form, but a server that parses the path as a form will read + as a space.
Why does encodeURI leave a colon and a slash alone?
encodeURI is for encoding a whole address, so it keeps the characters that give an address its structure: the scheme separator, the slashes, the query mark and the fragment marker. encodeURIComponent is for one piece of that address, so it encodes all of those except the unreserved characters, which is why a value meant to sit after ?q= needs encodeURIComponent.
What is the difference between 가 and 가?
Nothing but the base of the number. Both name the same character, U+AC00, the first Hangul syllable. The decimal form writes the code point in base ten and the hexadecimal form in base sixteen, and a browser reads both the same way. Named entities such as & exist only for a fixed list of characters, so everything else has to be written numerically.
Why do emoji need two \u escapes in JSON?
JSON strings are sequences of UTF-16 code units, and every code point above U+FFFF takes two of them. U+1F600 is written as the pair \uD83D\uDE00, the high and the low surrogate. In JavaScript source the newer form \u{1F600} does the same job in one piece, but it is not valid inside a JSON string and older parsers will not read it.
Is it safe to paste a token or a password here?
The page cannot send it anywhere: the content policy sets connect-src to none, the scripts are its own, and there is no server that could receive the text. It is not stored either; only the format and the options are kept in this browser. Still, a page is only as trustworthy as the copy you opened, so check the address before pasting anything secret.
Why does unescaping sometimes produce a replacement character?
Because the bytes did not form valid UTF-8. For example %EA%B0 alone is the first two bytes of a three-byte character, and a decoder has to put something in its place. The page shows U+FFFD, the replacement character, and says how many sequences failed, so the break can be found rather than hidden behind a guess.