Run this: db.foo.drop(); db.foo.createIndex({uniqField: 1}, { unique: true }); const docs = Array.from( {length: 10_000}, () => ({uniqField: Math .random()}), ); console.time( "insert non-dupe" ); db.foo.insertMany(docs); console.timeEnd( "insert non-dupe" ); console.time( "insert dupe" ); try { db.foo.insertMany(docs, { ordered: false }) } catch (e) {} console.timeEnd( "insert dupe" ); … and observe that the “insert dupe” is well over an order of magnitude slower, despite doing no actual writes: > mongosh `test-connstring.pl` dup_key_slow.js %s: %s insert non-dupe 195.07ms %s: %s insert dupe 8.752s