[tiensonqin]

Stored-session startup eagerly loaded preview/browser-authentication UI and an unopened sidebar. This pins LUI's on-demand preview and first-reveal drawer fixes, loads the public AuthenticationServices framework only for interactive sign-in, and keeps the full native navigation, controls, editing and authentication flows.

Corrects two launch-ordering defects: the ready root's opaque background was cleared on another main-queue turn, and the first-content observer could release sidebar loading behind that cover. Reveal now joins the ready root transaction. Sidebar loading waits for a display timestamp corresponding to an uncovered commit. The independent noninteractive logo fade remains unchanged; no timers, prewarming, private runtime overrides or UI omissions are used.

Includes isolated native regressions and an actual-AppMain fresh-process benchmark with the shipping FFI exports, optimized production modules, immutable binary checks, real graph titles and restored authentication. Profiling/CPU-counter observations are kept separate from uninstrumented performance samples.

Validation on current main 651202c:

• Previously rebuilt OCaml simulator core passed 1,090 tests; native link-input fingerprint 4e5bcf1282b9d0569d086ab3c5964dc869608e97e158f17c8cbdd3a24ccbb21b. This follow-up does not change core source.
• All 104 direct Swift Testing UI cases and eight native launch checks passed after the ordering fixes, including new actual-root first-transaction and covered-frame/display-clock regressions. New regressions were observed failing before correction.
• Production Release modules rebuilt without testability; dependency source checksums restored after every discarded prototype. No XCTest or swift test was run.
• Prior first-frame measurement regression and Release macOS LogseqModel gate passed.

Performance remains unresolved. Earlier seven-pair authentication controls measured 40.7 ms median pre-main saving but only 14.3 ms complete-frame saving. For the ordering follow-up, seven randomized three-way rounds show 64.5 ms paired window-to-content-commit saving, positive in every round, but only 1.9 ms complete-frame saving. An independent seven-pair confirmation shows 59.9 ms earlier content commit (7/7) and 11.6 ms complete-frame saving (4/7). Corrected complete-frame median in that confirmation is 707.1 ms; the 300 ms gate is still red.

Fresh profiles show approximately 34–36 ms of main-thread conformance checks, so the earlier 95 ms category is not a stable removable budget. Another observation finds substantial UIKit trait/layout execution after the earlier commit. Neither background comparison traversal nor an earlier commit can be counted as eliminated critical-path delay. Targeted drawer type boundaries, root tint omission, closed-drawer shadow/material changes, surface routing, padding and asynchronous-load prototypes failed causal controls or replication and were reverted. No experimental implementation from those controls ships.

This PR remains a draft. The extra first-layout cost is not fully explained or fixed. Uses fixes merged in logseq/lui#262, pinned at reviewed revision 3171aeaa. Investigation, rejected controls and raw measurements.