Write normal cases before testing limits
Choose a quick submission, a slower visitor and someone correcting a field before retrying. Define what counts as a new attempt versus correction. Include important language and mobile variants.
Fictional example: two genuine test users share a network address. Exercise their intended routes against the approved rule without assuming one IP always represents one person. Agree legitimate behaviour with an owner beforehand.
Let verification expire and try again
If verification results expire, leave the form open past the actual provider lifetime in the test environment. Check feedback and renewal. Do not invent the same validity period for every supplier.
Test correction after refusal and any reuse of an already processed test result under provider documentation. An old result must not produce false success, while visitors need to understand how to make a valid fresh attempt.
| Exercise | Question | Accepted when |
|---|---|---|
| Ordinary valid contact | Does normal use succeed? | Truthful confirmation and verified receipt |
| Slow visitor | Can expired verification be renewed? | Useful help without false delivery |
| Corrected field | Can the enquiry finish after correction? | Retry follows the intended route |
| Boundary in test environment | Do counting and recovery match policy? | Refusal and return to normal can be explained |
| Unavailable verification | Is outage distinguished from refusal? | Accurate state and working fallback |
Exercise limits and outages without volume testing
Use a dedicated low test threshold or simulated counter where the owner can do so safely. Test agreed triggering and recovery rather than loading the real service. Record which component returns refusal.
Then simulate verification failure. The interface must not confirm delivery when no enquiry was recorded. Preserve appropriate input for correction under the chosen design without placing sensitive text in broad logs.
Follow fallback contact and report differences
Have a test person use the intended fallback from the error message. Check visibility, mobile use and keyboard understanding. A decorative contact reference without a working destination is insufficient.
Report rule version, cases, observed refusals or receipt, owner and retest. A rejection count alone does not establish how many genuine customers were blocked. Follow up real complaints and measurement limits after an approved implementation.