Garbage collection and frame gaps
These separate 180-second web and desktop runs measure instrumented timing. Each role's GC and gap intervals use its own proved page clock.
Web
Role A: 127 merged main-thread GC intervals total 93.217 ms. The journal records 1 interval of at least 25 ms. The scoped table includes 0 and excludes 1.
Role B: 197 merged main-thread GC intervals total 201.498 ms. The journal records 404 intervals of at least 25 ms. The scoped table includes 398 and excludes 6.
Open the web figure at full size.
Role A is the source. Role B is the experimental direct-RGBA receiver. Both are research fixtures. Interval marks have a minimum visual width of 0.35 pixels, about 63 ms on a 180-second axis. Short marks are expanded. Their drawn width does not show their exact duration.
Result SHA-256: 38c082cc8f56dcc4321b04423070e996d198d4fc4e4b45b6d4865aa442e0de26
Desktop
Role A: 129 merged main-thread GC intervals total 99.715 ms. The journal records 0 intervals of at least 25 ms. The scoped table includes 0 and excludes 0.
Role B: 167 merged main-thread GC intervals total 174.091 ms. The journal records 430 intervals of at least 25 ms. The scoped table includes 428 and excludes 2.
Open the desktop figure at full size.
Role A is the source. Role B is the experimental direct-RGBA receiver. Both are research fixtures. Interval marks have a minimum visual width of 0.35 pixels, about 63 ms on a 180-second axis. Short marks are expanded. Their drawn width does not show their exact duration.
Result SHA-256: 49a2e8da3427c03805039a919b20581d412aa18a047d7992ef3e23ca1461611e
Method and limits
The scoped tables include recorded intervals of at least 25 ms wholly inside the first and last trace anchors. Pause absence checks also use recorded intervals that cross an anchor. Intervals from different stages can overlap. Their counts and durations must not be added as delivered-frame losses.
Only source-pinned MinorGC and MajorGC phases on the proved renderer main thread contribute pause intervals. An overlap shows correlation. It does not establish causation. A brief pause can coincide with a missed deadline. No overlap does not exclude other GC work or scheduling delays.
Memory samples use a separate controller clock. The plots do not align memory with page events. Heap and backing storage are sampled stocks, not allocation rates or packet-only storage. The shared native camera was active, but B received direct RGBA packets. These runs do not certify external virtual-camera consumers, sustained throughput or aging.