Reporting-table performance

Status and scope

Field	State
Contract status	Current master
Implementation status	Mixed; historical designs require remeasurement
Last source review	2026-08-26
Runtime verification	Required for performance and interaction claims
Owner	Not recorded in this public document

This master combines the C1/C2 design and hierarchy-table virtualization plan. It owns performance behavior for the paid-acquisition table without weakening frozen columns, interactions, or tenant scope.

Problem model

Historical investigations isolated redundant focus/expand fetch work and large nested renders, plus column-drag state invalidating a wide cell tree. Memoized row boundaries, stable handlers, debounced focus work, and cached derivations come before virtualization. Re-measure current source before reusing historical component counts.

Invariants

• Sticky columns keep offsets, z-index, widths, and horizontal-scroll behavior.
• Expansion, selection, sorting, visibility, resizing, reordering, and drill actions remain intact.
• Nested rows maintain stable identity and effective expanded state.
• Data fetching stays scoped and does not multiply by rendered row count.
• Performance work does not replace unavailable/missing values with zero.

Change sequence

1. Reproduce on the current public or authorized preview deployment.
2. Measure renders, queries, visible rows, and interaction latency.
3. Remove duplicate fetches and unstable state/handler dependencies.
4. Memoize the narrowest stable row boundary.
5. Cache expensive pure derivations with tests.
6. Consider windowing only after browser/component coverage exists for sticky and nested behavior.

Virtualization contract

A proposal must define the flattened visible-row model, stable keys, height policy, sticky-column composition, keyboard/focus behavior, nested expansion, scroll restoration, representative browser tests, and rollback without changing data contracts.

Acceptance gates

• Focus/expand produces the intended request count.
• Drag/resize stays within a measured render budget.
• Long names, hover disclosure, sticky offsets, and horizontal scrolling work.
• Nested interactions and tenant tests pass.
• Browser measurements are recorded; a local build is insufficient.

Provenance

Replaced document	Contribution retained here
C1/C2 table performance design	Memoization/debounce, frozen-column and rollback contracts
Hierarchy table virtualization plan	Hot paths, handler stabilization, staged windowing design