Seorang pengguna mengeklik sebuah tombol. Permintaan tertahan. Sepuluh detik keheningan. Mereka mengeklik tombol cadangan. Kini, dua pekerjaan dijalankan.

A user clicks a button. The request stalls. Ten seconds of silence. They click the fallback button. Now two jobs are running against a single intention. You end up with duplicate side effects, double charges, and a data mess that eats your afternoon.

This is not a frontend bug. A disabled button or a debounce timer in React will not save you. The first request was already in flight. The network simply swallowed the response. If your backend treats every incoming request as a brand new instruction, retries become liabilities. You need to fix this in your API design and your database schema.

The solution starts with a simple structural split.

Split Jobs from Attempts

Think of a job as the durable record of what the user wants. It captures the owner, the parameters, the target provider, and the exact intent. An attempt is a specific try at fulfilling that intent.

Picture a print shop. You hand over a file and they give you ticket #45. That ticket is the job. The shop tries the inkjet printer. It jams. That is attempt one. They move the file to the laser printer. That is attempt two. Throughout the process, ticket #45 never changes. If the shop issued a new ticket for every printer they tried, you would pay three times and receive three unwanted copies.

Your database should mirror this. One table holds jobs. Another table holds attempts. The job row stays constant while the attempts accumulate beneath it.

This separation gives you control. It also gives you a place to attach an idempotency key that survives network blips.

Require an Idempotency Key on Every Job

Every POST request that creates a job must carry a unique idempotency key. This key belongs to the user, not the session. Combine the owner ID and the key, then enforce a unique database constraint across those two columns.

Why a database constraint? Because checking for existence in application code before inserting is a race condition waiting to happen. Two identical requests can slip through the same microsecond gap. Let the database be the enforcer. If a user sends the same owner ID and key twice, the second request catches the unique violation and you return the existing job. Both requests get the same job ID. No duplicate work starts.

Be strict about scope. If someone reuses the key but changes the input payload, return a conflict. The idempotency key must bind to an exact intention, not just the user. Same key with different input means the client is confused, and your system should reject it rather than guess.

Protect State Transitions

An attempt is a state transition, not a new job. Your API must refuse to spawn a fresh attempt if a previous attempt is still hanging in a starting or unknown state.

Timeouts are the reason. When a provider request times out, the client sees failure, but the server-side process might still be alive. The GPU cluster could still be churning on your inference request. The container might still be writing to blob storage. If you mark the timed-out attempt as failed and immediately fire a second attempt, you are gambling with duplicate side effects.

Treat a timeout as an unknown state, not a failed one. Block new attempts until the earlier one reaches a terminal state or is explicitly cancelled by an out-of-band process. This pause is uncomfortable. It forces the user to wait. It also prevents the chaos of two workers mutating the same downstream resources.

Resolve Races with Compare-and-Swap

The hardest problems show up when multiple attempts finish. Maybe your system fired attempt one against the primary provider. After ten seconds of silence, it fired attempt two against the fallback. Now both attempts are done. You cannot let both write their results to the same job row.

Use compare-and-swap logic. Add a version number to the job row. When an attempt finishes, it runs an update with conditions:

• The current version must match what the attempt read at the start.
• No other attempt must have already claimed the result slot.
• If both pass, write the result and increment the version.

In SQL terms, that looks like an update statement with a WHERE id = $1 AND version = $2 AND completed_by IS NULL. If the update returns zero rows, another attempt already won. The late arrival must be ignored. Drop its result. Do not merge. Do not append. Throw the work away. A late result that overwrites an earlier winner is data corruption, and the only safe move is to discard it.

This handles the reverse-order finish cleanly. Attempt A leaves first but returns after thirty seconds. Attempt B leaves second but returns after five seconds. Attempt B wins the compare-and-swap. Attempt A’s update touches zero rows. Your system logs the race, ignores the stale payload, and moves on.

Test the Breakpoints

You will not catch these bugs in happy-path testing. Your suite needs to target the fractures.

• Simulate a double-click. Two simultaneous POST requests with the same idempotency key must return identical job IDs.
• Send the same key with mismatched input. Expect a conflict response. The system must not silently return the existing job if the parameters differ.
• Provoke a timeout. Verify the job lands in an unknown state, not a failed state, and that the system blocks further attempts until the ambiguity clears.
• Force two attempts to finish in reverse order. Confirm that the second one to return loses, even if the first one to leave was the official primary provider.

These tests are not edge-case luxuries. They are the contract your API makes with the rest of the system.

Validate Provider Intent Before You Fail Over

If you run a multi-provider setup, you might be tempted to treat different AI models as interchangeable slots. They share the same code path, the same HTTP client, and the same JSON schema. That does not mean they behave the same.

One model might hallucinate a top-level key. Another might ignore your system prompt formatting. Schema validation catches syntax errors, but it will pass a response that your business logic cannot interpret. A provider might return valid JSON that simply does the wrong thing with your prompt template.

Run provider-specific tests before you allow automatic model switching. Confirm that the fallback model actually respects your output structure at low temperature. Verify that your prompt renders correctly through that provider’s tokenizer. Test the full round trip with real inputs. Automatic failover is only safe when you have proven that the fallback shares the same operational contract.

Keep One Job Per Intent

Fallback paths are good. Uncontrolled fallback multiplication is a bug. Every layer of your stack needs to evaluate whether it has already seen the exact task. The load balancer, the API handler, the database, and the worker must all respect the same identity.

Build your system so that retries and fallbacks surface as new attempts under one stable job. Lock the job down with a database-backed idempotency key. Guard the transitions. Race the attempts. Let exactly one win. That is how you keep a single user click from turning into a weekend of data cleanup.