My Favorite Bugs: Invalid Surrogate Pairs
A rare and elusive bug in a collaborative text editor caused unsaved content to be silently lost during editing, with no clear error messages or reproducible pattern at first. The issue was eventually traced to specific emoji characters that use surrogate pairs in UTF-16 encoding, where inserting certain multi-byte emoji adjacent to each other would split a surrogate pair incorrectly. This corruption disrupted the underlying CRDT synchronization mechanism, leading to data loss upon reload.
- ▪The bug occurred in a collaborative editor using TipTap and Yjs, where edits would stop syncing silently, resulting in data loss.
- ▪It was triggered by inserting certain emoji above U+FFFF, which require surrogate pairs in UTF-16, at specific byte offsets that split the pair.
- ▪JavaScript operates on 16-bit code units by default, meaning string operations like .slice() can inadvertently break surrogate pairs if not handled carefully.
- ▪Only specific emoji and edit patterns caused the issue, making it extremely rare and difficult to reproduce.
- ▪The problem was identified when a product manager noticed it consistently occurred when editing sequences of red and green circle emoji used in status reports.
Opening excerpt (first ~120 words) tap to expand
My Favorite Bugs: Invalid Surrogate PairsMay 14th, 2026 • ~2,000 words • 8 minute readIf you're in the business of building things that run on computers long enough, I think you will eventually acquire a favorite bug story. This is a short story about mine. I've also built an interactive tool where you can explore the concepts underpinning the heart of this bug.The bug: two emoji enter, none leaveI was working on migrating a legacy editor to a more collaborative experience with my team. TipTap on top (itself a wrapper around ProseMirror), Yjs underneath handling the CRDT magic for real-time syncing. It worked well! Mostly.In our alpha/early release days, when it was still mostly internal and/or early rollout users, sometimes the editor would just stop saving your content. Silently.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at George Mandis.
Discussion
0 commentsMore from George Mandis
