Benchmarks

Small check: under synthetic bursts faster than the demo simulator (~0.5 evt/s), the capped ring still stops at MAX_EVENTS (10 000).

What this is (and is not)

npm run bench runs runAppendBenchmark in Node / Vitest.

It checks	It does not claim
Buffer length stays ≤ 10 000 after bursts	Browser FPS or React commit cost
Append path stays usable for synthetic 100 / 500 / 1000 evt/s payloads	Production load or multi-tab stress
Approx byte estimate is a rough lower bound	Accurate heap profiling

Treat this as a performance consideration for the data structure — not a scalability guarantee.

How to run

npm run bench

Representative results

Run on a developer machine while writing this page. Absolute timings vary; the useful invariant is final buffer length ≤ 10 000.

Target evt/s	Events in the 1s window	Final buffer length	Cap
100	100	≤ 10 000	10 000
500	500	≤ 10 000	10 000
1000	1000	≤ 10 000	10 000

Why these numbers matter for the UI

• The ring overwrites oldest events when full — history is bounded on purpose.
• The worker only receives a short sample window, not the whole buffer.
• appendEvents exists so tests can publish batches without one React snapshot per event.

More on the buffer: Decisions & challenges · Pipeline