This tests a fork of Automerge without the use of Object.freeze. This fork is located under node_modules/hot-automerge. Run with yarn start or npm run start.
After removing calls to Object.freeze from automerge (hot-automerge), testing 10,000 inserts on my 2012 Retina MacBook Pro (2.3 GHz i7, 8GB RAM) I get:
automerge profile 10000 inserts: 55546.893ms
hot-automerge profile 10000 inserts: 4051.241ms
which gives 55s using Object.freeze vs. 4s without it—a ~90% performance improvement. Repeated runs gave similar results, as did changing the order of the tests.
A very rudimentary test on the performance impact of Object.freeze is available on JSPerf here. JSPerf gave a roughly 42-47% slowdown using a single Object.freeze call.
In this case, multiple calls are made per-insert, causing a greater performance penalty.