Bound runtime resources and fix debugger, validation, and mobile controls

[rendis]

Summary

Unbounded invoke launches, retained event/history data, and synchronous callback reentry could exhaust resources or block execution. Add configurable runtime budgets, a shared invoke manager, cooperative cancellation/shutdown, cancellable lock waiting, and context-aware reentry checks. Zero-value options preserve existing policies and mandatory interfaces remain unchanged.

• Bound invoke concurrency across machine constants, universe constants, and reality/transition invokes; optionally cancel work on exit, restoration, or positioning. Add lifecycle and contextual snapshot capabilities.
• Limit accumulator entries, tracking, and bot history; preserve bounded history during rollback. Add optional strict observer registration checks and propagate bot snapshot/cancellation errors.
• Fix CLI reflection panics and empty selections, and test real-machine event history, rollback, and checkpoint loading.
• Return structured errors for malformed Studio definitions and resolve references through own properties, preserving explicitly declared names such as constructor.
• Support multiple Web Component tag names; cover attributes/properties, mounting, reconnection, and DOM events. Fit header, search, machine panel, and bottom controls on mobile.
• Expand and complete mutation campaigns; fix incompatible Babel overrides and exclude mutation sandboxes from normal test discovery. Add a reproducible history benchmark.

Validation

GitHub Quality (Go and Studio) and CodeQL passed at 6e0bd69a553dce9e4a395bf087ffd0b9fdb54221. The first CI fuzz run ended with a runner context deadline exceeded and no failing input; a local repeat and the CI rerun passed without code changes.

• Go build/vet and race-enabled tests with coverage passed across all 11 packages; three native fuzz smoke targets passed.
• Studio frozen installation, lint, typecheck, 283 tests, all builds, and dependency audit passed; zero advisories including development dependencies.
• Complete Gremlins campaigns: builtin 92.00%, bot 96.00%, root 83.11%, experimental 87.31% efficacy. The root campaign preceded the final private async-context adjustment; final experimental mutation and race tests cover the resulting runtime.
• Complete Stryker campaign: 845 mutants, 59 tests, 58.11% total / 65.12% covered score; 489 killed, 2 timed out, 263 survived, 91 uncovered, no runner errors.
• Chromium: header/search/control bounds at 320, 390, 768, and 1440 px; library and JSON export; production layout at 320 and 1440 px; actual Web Component locale/event flow. CLI statement coverage increased from 1.3% to 51.0%.

Contracts and measured costs

Cancellation requires cooperating user callbacks. Legacy context-free owner snapshot calls inside synchronous callbacks can still deadlock; use callback arguments or the new contextual APIs while preserving the supplied context. Admission/cancellation failures do not roll back prior side effects.

ELK remains approximately 1,458 kB (445 kB gzip) and is requested by automatic initial layout despite being a separate chunk. Full checkpoint copies preserve isolation from consumer mutations; at 5,000 realities the benchmark measured about 53 ms per undo/redo pair. These costs and mutation survivors remain documented follow-up opportunities.

Runtime usage · Review, mutation counts, and benchmarks