Define the final condition and reset state
Identify the constrained resource and what a confirmed booking means. If the last place is only a manually assessed request, test that model. A temporary hold and final registration need different expectations.
Fictional example: a course has one test place left without affecting real participants. Two separate sessions load the same option before either submits. Record starting capacity, test accounts, expected result and how trial bookings will be removed.
Make the attempts compete at the decision point
Send the attempts so they compete for the same capacity. Two tabs booking far apart in time are a different test. Record each attempt reference and what actually reached the capacity system.
Check that the rule is enforced by the authoritative function. A database solution may need transactions and appropriate concurrency control; an external booking provider has its own rules. PostgreSQL documents different properties of isolation levels. A browser button setting cannot replace that control.
| Trial | Check in the source system | Check for the visitor |
|---|---|---|
| Two attempts, one place | The capacity rule holds | Only the agreed number get confirmation |
| Same attempt repeated | No unintended additional booking | Same outcome or a clear investigation |
| Hold expires | Capacity released under the model | An old hold is not shown as valid |
| Response lost after registration | Actual status can be investigated | No invented success or failure |
| Trial booking cancelled | Correct capacity after withdrawal | Message matches actual status |
Retry without creating another business decision
A network failure can occur after successful registration. Decide how a visitor or support owner checks the existing attempt before creating another. A new random reference on every retry can hide that the same booking was intended.
Follow the integration's documented identification and repetition rules. An HTTP method name does not establish that the booking logic handles the same attempt safely. Include a genuinely new booking too, so duplicate protection does not block legitimate additional reservations.
Reconcile capacity at the end
Compare final capacity, valid holds and confirmed bookings with expectations. Read messages in both sessions. One correct database row is not enough if the unsuccessful visitor receives false confirmation or lacks an understandable next step.
Keep version, times, references, deviation and retest. Reset the trial data and check that no notifications reached actual customers. The trial establishes the agreed model in the test environment; production needs its own controlled verification.