I4200 — A warm UI screen borrows what its sources answered and writes its page once, so a plan of any size is not copied per request

○READY wave 0 · p1 · implementation profile · runs alone

Part of M004 — Performance is executable evidence, and the hot path does no canonical work twice.

Objective

Make the warm request path of /ui copy nothing it already holds: read the intents, issues, waves and scope pairs through view::items (borrowed) instead of view::list (a deep copy) in ui/issues.rs, ui/idd.rs (screen, render, choice, wave_panel) and ui/intent.rs's inspector; move a screen's tree into its page in ui/shell.rs::page instead of cloning it; and escape text and attribute values straight into the page buffer in cockpit/html.rs instead of through a new String each.

Why

Train 5's ui-checks step 'every warm screen of the UI is within its budget' failed on #2403: system.http.ui.idd warm p95 108.3 ms and system.http.ui.issues 106.6 ms against 100 ms, and both sat at 95-110% of the budget on every runner. The data comes from cached readings, so warm cost equalled cold cost: every request deep-copied the whole intents list (four times on /ui/idd), every issue of the plan, the whole waves answer with every scope pair of the plan (79% of /ui/idd's render once the first copies were gone), and then the screen's entire element tree again to place it in the page, which on /ui/issues is a ~970 KB page of 976 rows.

Current state

Measured on d078fb2309e, both builds made the same way here (debug, RUSTFLAGS=''), `majordomus bench <id> --transport system --profile ci`, warm p95: system.http.ui.idd 17.9 ms and system.http.ui.issues 35.8 ms. view::list is `items(answer, pointer).to_vec()`; shell::page took &View and did `.child(view.main.clone())`; cockpit::html::El::write called escape(), which returns a new String, for every text and every attribute value.

Desired state

Warm p95 on the same build: system.http.ui.idd 6.4 ms (-64%) and system.http.ui.issues 22.2 ms (-38%), with every screen's output unchanged byte for byte, so both screens sit well inside the 100 ms budget on CI's runners without raising it.

Scope

• .ai/repo/project/issues/I4200.yaml
• apps/majordomus-cli/src/ui/issues.rs
• apps/majordomus-cli/src/ui/idd.rs
• apps/majordomus-cli/src/ui/intent.rs
• apps/majordomus-cli/src/ui/shell.rs
• apps/majordomus-cli/src/ui/mod.rs
• apps/majordomus-cli/src/cockpit/html.rs

Out of scope

• Raising any warm budget
• Paginating /ui/issues, which would change what the screen shows

Dependencies

None. This issue is a root of the graph.

Acceptance criteria

• The ui and cockpit::html unit tests pass unchanged, so every screen renders the same bytes
• bench system.http.ui.idd and system.http.ui.issues warm p95 fall by at least a third against a build of the same commit without the change
• ui-checks' warm-budget step passes on CI without a rerun

Validation

• cargo test --manifest-path apps/majordomus-cli/Cargo.toml --lib -- cockpit::html ui::
• majordomus bench system.http.ui.idd --transport system --profile ci --no-write
• majordomus bench system.http.ui.issues --transport system --profile ci --no-write

Evidence required

• warm_screens_copy_nothing
• warm_budget_passes_on_ci

Evidence

None recorded. Every token above needs a command or an artifact behind it before this issue can be completed; narrative is refused.

Timeline

started

—

verified

—

completed

—

Those three fields, the evidence above and the state of the dependencies are all the status is made of. There is no status field to disagree with them.

Canonical record: .ai/repo/project/issues/I4200.yaml. Read it back with majordomus plan show I4200.