A string split at an arbitrary index can cut a surrogate pair, and each half is sent as U+FFFD

Fact

JavaScript strings are UTF-16, so a character outside the Basic Multilingual Plane (most emoji, many CJK extension characters) is two code units. Slicing at a fixed length, as chunking for a size limit does, can end one piece with the first half and start the next with the second. WebSocket.send, RTCDataChannel.send and TextEncoder all take a USVString, so each lone half is replaced with U+FFFD before it is encoded. The reassembled text has two replacement characters where the emoji was, and nothing throws.

Why it matters

The corruption is rare and content-dependent: it needs a non-BMP character to straddle a piece boundary, so it passes every test with ASCII fixtures and shows up as a mangled name or a JSON parse error only for some users. It looks like a decoding bug on the receiving side, whose code is in fact correct.

How to apply

• When cutting at index end, step back one if the unit before it is a high surrogate:

  const c = text.charCodeAt(end - 1);
  if (end < text.length && c >= 0xd800 && c <= 0xdbff) end--;

• Or encode first and split the bytes, sending binary frames.
• Test chunking with a payload made of emoji and a piece size that is odd.