[chriskinal]

The 2026-11-14 nightly reported VALVES-NOT-FOLLOWING in J19 (1.1 s against the 1 s limit). That night the rig runner's emulated machine module received 0.6–4.5 % fewer PGN 239/229 frames than the night before in 12 jobs, J19 among them (606,522 of 635,300). The other job harnesses on the same build received every frame. The module hears the app over real UDP on its own thread, while the run goes faster than the clock.

Change

• After each 0.1 s fix, the frames sent during it get up to 5 ms of wall time to reach the module (JobRun.WaitForFrames).
• ValveFollowTally counts the PGN 239 frames the module never got while online.
• A screen/valve mismatch over the 1 s grace in which frames were lost is counted apart and reported as a harness note with the count; it no longer counts against AgOpenWeb.
• A module that is off or whose Wi-Fi is down loses frames on purpose, so those mismatches still count against the app (C7).
• README row updated.

Tests

• ValveFollowTallyTests (5, run without a rig, in milliseconds).
• Runner, J19 + C7, C8, C9, C16, whole jobs, this branch against develop: 0 new / 0 gone in both reports.
  • Base: the module missed frames even on the Runner (J19 631,803 of 635,300; C7–C16 up to 37 each).
  • Fix: every frame arrived (J19 635,300), so no harness note was needed.
  • C7 and C8 still report VALVES-NOT-FOLLOWING as expected.
  • Run time 19 m 29 s → 19 m 46 s.

🤖 Generated with Claude Code