The user put K-Mix on the same Satechi hub with WRLD.BLDR and PD and explicitly retained disabled app audio inputs. SmartGrid PID 22946 was launched at 00:01:12 PDT with normal DSP, visible UI, autoplay, MIDI sender/SysEx and remaining workers enabled; no source or installed binary changed. Actual K-Mix route, 48 kHz/512 frames, zero app inputs and seven active app outputs were verified. The native session exposes ten outputs, versus MAYA’s four: this is an inherent configuration difference, not an equal-channel control. Initial thermal state was nominal; external power present and battery charging off. No induced conditions were active.
K-Mix reproduced kernel endpoint 0x82 status 0xe00002ed transaction errors, AUAInputTransferManager failure handling, I/O restarts and approximately 115–125 ms app callback gaps. Examples in the live log occur at 00:03:29.258795 and 00:03:35.138606. At 00:03:46.073252, endpoint 0x81 also reported the same status and the feedback reader logged use of its previous value. These are transport observations, not measurements of audible output; no analog recording was available. MAYA is not necessary for this reported failure chain. Final event counts and one-to-one clock correlation remain pending.
The first collector saved /private/tmp/smartgrid-ipad-kmix-logs-20260911-000111. Its startup gate read only twelve callbacks because app initialization took approximately thirteen seconds; it ended live capture rather than treating an insufficient startup sample as verified. It left the app running and collected an archive. This is a capture verification failure, not evidence of an app startup failure. Its initial stream included historical/debug messages and has substantial observation overhead.
The second collector attached to the same process without relaunching at 00:03:07; verification passed at 00:03:09.596755. Raw files are /private/tmp/smartgrid-ipad-kmix-logs-20260911-000306. Live system collection uses payload/info/no-sensitive flags, excluding historical replay and debug/callstack expansion; relevant system records are saved with timestamps. App log is copied incrementally every fifteen seconds, with one-minute battery snapshots. This Wi-Fi traffic is a measurement condition. Full app history since launch is preserved in app.log.
At 00:04:40 the app reported its output route changing back to MAYA. The capture guard stopped at 00:04:42.325; the user then explicitly requested stopping recording to try something else. Exclude the route transition and subsequent MAYA activity from K-Mix measurements. The status value capture_failed reflects the route guard, not a claim that the recorder lost the preceding evidence. Final archive collection succeeded at 00:05:38.317 and the device connection is closed. Playback was left running. No heartbeat was created; an attempted creation returned a missing destination argument, and the user’s stop made it obsolete before retry.
The user is trying another configuration. Do not restart capture, relaunch or alter their setup without their next instruction. Machine-readable checkpoint: /private/tmp/smartgrid-research-20260909/wireless-experiment-status.json, kmix_substitution.
Archive/app analysis completed after the user stopped the next run. K-Mix playback covers approximately193.53seconds, 00:01:25.849–00:04:39.380. It contains23kernel transaction errors:20on input endpoint0x82 and3on feedback endpoint0x81. All20input errors uniquely match20audio-server restarts and20callback gaps. Gap minimum/median/maximum105.128/124.502/133.722ms. The three feedback errors each have a use-previous-feedback-value report; there are no three additional restart episodes. All failures use0xe00002ed. After initialization, thermal remainsnominal, callbacks512at48k, no measuredprocessingoverruns orcallbacksequenceholes.
Exclude31additional transaction-error messages associated with disconnect (27endpoint0x83,2endpoint0x81,2endpoint0x82). Kernel explicitly records K-Mix hardware connection lost at00:04:39.476 in the archive; the simultaneous fault burst begins just before that. This is excluded from playback failures. Input errors had already recurred well before disconnect.
The next experiment’s physical configuration has not been explained by the user. Initial attach-only segment /private/tmp/smartgrid-ipad-adhoc-logs-20260911-001809 started00:18:11 and was ended for a user-requested normal-mode relaunch; its live/app logs are preserved, automatic archive was canceled to release the pairing promptly. Do not combine its timeline with the relaunched run as a single uninterrupted exposure.
The requested normal DSP/UI/autoplay/MIDI launch occurred00:22:04.625, PID23132. Actual MAYA48k/512, zeroappinputs, nominalthermal verified. Logs are in /private/tmp/smartgrid-ipad-adhoc-logs-20260911-002203. Stable playback begins00:22:10.973 and lasts approximately154.62seconds before the interaction interval: zero transaction errors, zero restart-related callback gaps and zero measuredprocessingoverruns after initialization. The initial470frame callback overran; subsequentcallbacks settle512.
One input transaction error occurs around00:24:45 (live00:24:45.658501, archive00:24:45.742308), with restart and a callbackgap. User explicitly confirmed changing/stopping something at that moment; classify it as interaction-associated and exclude it from spontaneous-playback counts. App entered background around00:24:46.089 and callbacks ceased. Do not count idle recorder time afterward as additional clean playback. Log capture stopped00:25:53.722, finalarchive succeeded00:26:42.860; connection closed and no new experiment started.
Use archive counts without adding duplicate live records. Live/archived wall timestamps differ by roughly57–84ms in these captures, so submillisecond causal intervals are not established by cross-source walltime. Callback duration is measured in the app monotonic domain; correlation is one-to-one within bounded windows. No analog recordings exist for either run; absence of periodic audible corruption is not established. The change explains a short quiet window only, not a proven fix. Counts and individual matches.