Pending describes unfinished work. Unknown describes a conclusion the available evidence cannot establish. Failed describes an operation or check that did not succeed according to its own contract. None of these labels automatically means the email was observed in Spam. Read the object and stage attached to a state before deciding whether to wait, retry an observation, investigate transport, or correct a real technical failure.
Ask what the state belongs to
A report can contain a test-run state, a mailbox-access state, a message-receipt state, and individual diagnostic results. A failed DNS lookup and a failed SMTP authentication check concern different objects. A run marked complete can still contain unavailable evidence for one provider. A received copy can have a known folder while a particular diagnostic remains unknown.
Start by writing a short sentence for the state: the mailbox could not be read, the folder check has not finished, or the receiver reported a particular authentication failure. If the interface label cannot support such a sentence, inspect its detail or methodology. Do not let a single colour stand in for several different technical meanings.
Distinguish pending from a completed absence
While a check is pending, the observer has not completed the relevant work. A message may arrive later, but the pending label itself does not prove it was sent or delayed. Keep transport submission and retries in a separate timeline. Repeatedly pressing send while observation is pending can create duplicates without answering why the first check has not completed.
A successfully searched mailbox with no matching message supports a not-found result at that time. An inaccessible mailbox supports unavailable evidence instead. These are different from pending because a completed operation, or its inability to proceed, has already supplied a narrower result. Preserve those distinctions when exporting or summarizing the report.
Interpret diagnostic failures at their own scope
An authentication failure on a received copy is evidence about that check and identity. It does not by itself tell you the observed folder or the receiver's entire filtering rationale. Conversely, a DNS lookup timeout does not establish that the domain has no record. A transient network error requires a different response from a confirmed invalid record.
Look for timestamps, the checked identity, and the evidence source. A current DNS query may differ from what a receiver saw at send time. An AI explanation can remain uncertain even when several underlying facts are known. Keep the factual check, the operational error, and the interpretation in separate fields in your incident notes.
Choose the next action from the evidence
For pending work, inspect progress and allow the documented operation to finish. For unavailable observation, check service health or permitted access and retry the read operation where appropriate. For a completed not-found result, investigate submission, address, matching, and transport. For an established authentication defect, correct the relevant sending configuration and test a fresh copy.
Retain the original state and timestamp when a later check resolves it. A changed state supplies a history, not proof that the earlier report was dishonest. When comparing runs, use similar stages and show unresolved counts beside percentages. Unknown results should not be silently discarded if doing so changes the apparent coverage or confidence of the conclusion.
Frequently asked questions
Should unknown be displayed as a failure?
It should remain distinguishable. An unknown conclusion and a demonstrated defect require different decisions, even when both deserve attention before release.
Can a completed run still have pending details?
That depends on the product's state model. Read the particular field's contract and timestamp. Do not infer that every subcheck completed merely from a high-level label.
Is retrying a check the same as resending?
No. Retrying observation of an existing copy and sending a new copy are different operations. Record which action occurred so later evidence remains attributable.
Sources and scope
Source check: November 5, 2026. Source checking is separate from publication. Observations apply to the tested copies and conditions.