Start here

the test fails one run in twenty, everyone reruns CI until it goes green, and the suite stops meaning anything.

Flaky testBackend

you went to click the button, an ad loaded above it, and you clicked something else entirely.

Cumulative layout shiftFrontend

the meeting spent five minutes on the data retention policy and thirty-five on the button label.

BikesheddingArchitecture

there is one machine everyone is afraid to reboot, and its configuration exists only in the history of who logged in.

Snowflake serverPlatform

every task is green, the run took its usual eleven minutes, and it processed zero rows.

Silent failureData

you sent the same prompt twice and got two different answers, and nothing in your code changed between the calls.

Non-determinismAI

there is a 3 in the middle of the expression and nobody left a note saying which three that is.

Magic numberPatterns

nothing is broken, but the way the parameters keep getting passed through five functions feels wrong.

Code smellArchitecture

you changed the function, everything passed, and it turned out nothing had called it for a year.

Dead codePatterns

you changed one module and three tests in a directory you had never opened went red.

CouplingArchitecture

the code for one job is spread over four files, and each of those files also does three unrelated jobs.

CohesionArchitecture

the same value was stored in two places and they drifted apart, so the header and the sidebar disagreed.

Single source of truthFrontend

the vendor added three columns and renamed one over the weekend, and nobody told anybody.

Schema driftData

the class asks for what it needs in its constructor and never goes looking, so a test can hand it anything.

Dependency injectionPatterns

you changed the shape of the code all afternoon and the tests never went red once.

RefactoringArchitecture

you shipped the quick version to make the date, wrote down what you skipped, and paid for it every sprint after.

Technical debtArchitecture

you measure the delay before the work starts moving, separately from how long the work itself takes.

LatencyBackend

you changed the row and now you have to hunt down every cached copy of it before someone reads the old one.

Cache invalidationBackend

you fetch a list of fifty orders and then the ORM quietly issues fifty more queries, one per order, to load the customer.

N+1 queryBackend

the bug only shows up under load, never in tests, and the outcome depends on which of two threads got there first.

Race conditionBackend

you can safely run the same operation five times and the system ends up in exactly the state it would after one.

IdempotentBackend

you give up on the call after two seconds instead of holding the thread open until the other side eventually answers or does not.

TimeoutBackend

you retry the failed call, wait twice as long, retry again, and add a little randomness so your whole fleet does not retry in lockstep.

Exponential backoffAI

after enough consecutive failures you stop calling the provider at all for a while, instead of queueing thousands of doomed requests behind a dead dependency.

Circuit breakerAI

every bill, rate limit and context error you will ever see is denominated in tokens, not words or requests.

TokenAI

everything you send and everything it says has to fit one budget, and the retrieved documents are what usually blow it.

Context windowAI

you asked for a citation and got a real-looking one that does not exist, delivered with the same confidence as the parts that were true.

HallucinationAI

you put the new build on the servers, which is a completely separate event from anybody actually using the new behaviour.

DeploymentBackend

you flip the flag and users start getting the new behaviour, on code that has already been running in production for a week.

ReleaseBackend

you ship the code turned off, switch it on for yourself, then for one percent, then for everyone.

Feature flagBackend

the deploy broke something so you put the previous version back rather than trying to fix forward under pressure.

RollbackBackend

it worked in staging and broke in production, because staging had one replica, no mesh and a different database size.

Environment parityPlatform

you write up what happened focusing on why the system allowed it, not on who typed the command.

Blameless postmortemBackend