The OT library we use is pretty rough, it usually works fine but sometimes when documents get long there are pathological cases.
The plan is to shift to a CRDT based implementation, though I do not have a timeline yet.
「いいね！」 2