docs(examples): a jumps-and-merges flow for the layout examples

[ako]

What

Adds Pattern 6 — CxLayout.CX_SYNC_FactoryTelemetry to mdl-examples/doctype-tests/02c-complex-layout-examples.mdl. One flow, 37 elements, 4 merges, 11 joins. No code changes.

The file's five existing flows predate named merge / join, so nothing in it exercises the graphs a nested if cannot spell — which is exactly the part of the layout algorithm with no example coverage when that algorithm changes.

Pattern 6 covers all of them in one realistic flow:

Shape	How it appears
Backward jump	the java-action handler does join attempt, where merge attempt sits above the call — the retry edge runs right-to-left
Error → main flow	a handler that gives up does join degraded; the main flow reaches that same merge when the response is empty
Crossed branches	the inner split's FALSE arm and the outer split's FALSE arm both land on review
Fall-through handler	the last handler ends without return / join, merging back on its own
Funnel	three independent paths end at finish

Testing

• mxcli check … --references against a Mendix 11.14.0 MPR v2 app: passes.
• make check-mdl: passes, 02c included.
• exec then re-exec: Created then Unchanged — idempotent.
• Measured against fix: lay out microflows written without @position so they can be read #1155 (the layout PR) on that app: 2995 × 540 vs 4790 × 350 on current main, same 37 elements, flow identical apart from geometry.

Two things mxcli check caught in the first draft, both kept as-is in the final flow: an activity always creates its own output variable (declaring $RawResponse first is CE0111), and publish is a reserved word, so the merge label is emit.

For the reviewer

This example is also a repro for two gaps left in #923, written up with evidence in #923 (comment). Both reproduce on the pre-#1155 build, so they are pre-existing and unrelated to the layout work:

1. A one-inbound merge that exec writes and the next exec deletes — describe says so itself ("re-executing this MDL DELETES it"), and the warning is gone after the round trip. Reduced to a 25-line repro in that comment.
2. This flow goes from 0 describe warnings to 2 after a single describe → exec, the second being "This description is NOT equivalent to the microflow and must not be re-executed over it". Top-level merge sections get pulled inside the branch that joins them and one merge is dropped.

So: do not maintain this flow through describe → exec. It is the one flow in the file where that round trip is lossy, which is a property of the graph shape, not of this PR.

No issue was filed first (CONTRIBUTING.md step 1) — this came out of testing #1155 by hand. Happy to open one retrospectively if you'd prefer the trail.

🤖 Generated with Claude Code