Topic
A class of software defects where operations silently produce incorrect results instead of raising errors, letting bad data persist undetected.
n8n 2.41.6's Postgres node writes $1 values into SQL text via pg-promise 11.9.1, so Postgres never gets a bind parameter. A psql PREPARE dry-run tests a path the node never uses, and Postgres types inlined literals before any CASE guard runs.
Reality
open() and subprocess.run(text=True) both fall back to whatever GetACP returns, so a UTF-8 file read on an en-US machine comes back as cp1252 mojibake with no exception raised, and chcp 65001 changes a different setting.
Reality
Two weeks of field notes from an agent with write access to a construction company's books and bank account point the reliability work at verification, because every failure in the log reported success first.
Reality
The fourteen standard PDF fonts are Windows-1252 underneath, so a Polish company name throws an encoder error, and the one-line regex that stops the crash ships a valid invoice with the letters quietly gone.
Reality
Resolving a YouTube handle by regexing the first channelId out of the page HTML returned the wrong channel 15 times in a 16-channel sample, because the string it finds first belongs to whatever the page features.
Reality
Maintenance work where a wrong fix returns a plausible answer needs an oracle before it needs a model, and one production deployment now reports what happened when the model was given a single judgement and nothing else.
Reality
A minimum over positive bigints started from 0n, so every real loss lost to the seed. Nothing threw, and the assertion that would have caught it sits on a secondary field most suites never inspect.
Reality
A dev.to post swept 146 uses of IFS=$'\t' read in one codebase and found 14 live defects. The corrupt parse and a legitimate one have identical shape.
Reality
A sign convention chosen by the author sat on both sides of the invariant, so mutation testing could not reach it. Fifteen minutes with a real importer could.
Reality
A mutation audit on Engrava found guards that validated a value then discarded the result, and two hygiene tests that had been passing for reasons their own docstrings got wrong.
Reality
Seven Python parsers, 300 labelled malformed LLM outputs. On recoverable cases json-repair wins by two; count the 25 unrecoverable ones and it loses by seventeen.
Reality
SecretSpec's dotenv-ng 1.0 makes dollar signs literal after dotenvy silently dropped parts of a bcrypt hash. The damage surfaced as an authentication failure, far from the parser that caused it.
Reality
A worn or counterfeit microSD card can acknowledge a write and keep nothing. Free space drops, the file appears, and the gap only shows up on the day you go looking.
Reality
A dev.to writeup traces two weeks of low-looking referral counts to a @Modifying increment that never bumped @Version, letting one stale save() overwrite the counter.
Reality