Edoed: Sounds like a classic case of a phantom bug. First thing—dump the raw stream into a file and run a diff against a known-good baseline. That way you can isolate the exact cycle where the flicker appears. If you suspect a backdoor, try crafting a minimal payload that flips a flag you know is hidden; put it in a separate thread so you can see if the simulation reacts without touching the main loop. I once found a stray environment variable that was acting like a secret switch after a refactor; it was just a typo in a config file. Keep a versioned log of every tweak—Git will freak out if you try to commit the same line twice, but that’s a good reminder to test in isolation. If the test triggers, you’ve got a candidate. If not, keep hunting for a hidden parameter in the build scripts or the config parser. Remember, the trick is to keep the test isolated so you can pinpoint the exact moment the simulation changes state. Happy hunting!

That’s the vibe I was hoping for—keep the logs granular, maybe add a checksum of the state snapshot at each tick. If the clocks are fuzzy, try forcing a deterministic tick by stalling the main loop with a dummy task. Sometimes the race shows up only when the scheduler gives a burst to a low‑priority thread. Good luck, and if you find that hidden flag, send it my way—I’ll add it to the repo and pull request a fix.

Sounds like a solid approach. Just remember to keep the dummy task lightweight; otherwise it could introduce its own noise. If you hit a stray flag, I’ll squash it quickly—just ping me. Good luck with the detective work.