N8N · MAIN-BRANCH CONVERGENCE
One node.
More than one run?
Trace the branches that reach a shared step. Review an Append change and keep node executions, items and HTTP requests separate.
A shared node can run again when another branch reaches it. An Append Merge waits for its inputs, but five items can still generate five HTTP requests. Whether batching changes the bill depends on the next step and the API.
CONTROLLED NATIVE EXPERIMENT
Same five items. Three outcomes.
Executed with n8n 2.45.4, execution order v1, synthetic data and an ephemeral local HTTP counter. No paid API or production visitors were involved.
| Route | HTTP node runs | Actual HTTP calls |
|---|---|---|
| Two branches → batch Code → HTTP | 2 | 2 |
| Append → batch Code → HTTP | 1 | 1 |
| Append → HTTP, one item per request | 1 | 5 |
Inspect the native receipts and fixture hashes · Download the evidence kit
The rewrite generated by this tool was separately imported and executed in n8n: one batch run, one HTTP call, five items retained. The kit includes a loopback reproduction script; raw execution logs are excluded.
ANALYZE YOUR EXPORT
Find the first shared step
Download a workflow JSON from n8n, then choose a file or paste it. Supported scope: up to 500 nodes and 2,500 main connections, one trigger at a time. Separate Manual and Schedule triggers are separate executions.
Your file and node parameters are not uploaded. Example interactions are excluded from useful-action counts.
Your convergence review
Path counts are potential arrivals if branches emit data. Code, empty outputs, conditions, loops, retries, sub-workflows and API batching can change the result. This is a focused graph review; it does not execute your export.
Review the Append change
The original workflow is unchanged. The downloaded clone retains your parameters and credential references: keep it private.
Source and scope
n8n Merge documentation: Append and connected inputs. The native benchmark and generated rewrite were checked on October 27, 2026. Other versions and provider behaviors need their own test.
Workflow content is processed in page memory. First-party measurement sends only an experiment ID, action, coarse source, format, an ephemeral page-session ID, sequence and test marker. It receives no workflow JSON, code, parameters or report. Measurement privacy.
Created with AI assistance and controlled native tests. CompSetly is independent of n8n.