A DDoS retest works by repeating a failed run with every variable held fixed except the fix, rather than by running a fresh DDoS test against whatever the stack looks like today. It is the step that turns an open finding into a closed one, and it can only do that if the second run can be compared with the first.
What a DDoS retest holds fixed
The original DDoS test report should already name the exact vector, rate and scope that must run again to close each finding. A retest honours that condition to the letter, and a few variables need more care than the report usually gives them:
- Rate steps and hold times. A shorter hold can end a step before a slow failure appears, such as a connection table that fills over minutes. The step passes because it stopped early.
- Source spread. Suppose the fix was a per-address rate limit. A retest from fewer source addresses trips that limit sooner and reports a pass the original spread would never have produced.
- The legitimate-traffic model and the goodput threshold that defines failure.
- Measurement points and the clock, so detection and recovery times line up with the first run's.
- Allowlists. Generator ranges added to an allowlist while someone was troubleshooting turn the retest into a test of the origin alone.
Change any of these and the result is still useful, but it is a new test, not a retest.
Before it starts
Check that the fix is in the state the retest assumes. It is deployed on every hostname, region and edge configuration the original run touched, it is enforcing rather than logging (the block vs count mode trap catches fixes too), and its version is written down.
Then list what else changed since the first run. Every other change is a rival explanation for a different result. If the list is long, the honest label for the run is a new baseline.
How the run proceeds
First, repeat the step that failed, exactly as it ran before. That is the only reading that answers the question the finding asked.
Then keep climbing. Once one component stops being the weakest, another takes its place, so the layer of first failure found past the old step is the next finding rather than a surprise later.
Finally, give the vectors that held last time a short rerun. Fixes move load around. A stricter challenge rule shifts work onto the edge; a tighter rate limit starts turning away an API integration that never failed before.
Timing matters too. A retest run soon after the fix has few other changes to untangle. One run months later is mostly measuring drift, which is why a fixed failure appears among the triggers for DDoS testing frequency rather than waiting for the next scheduled test.
What a DDoS retest reports
Each finding ends in one of three states:
- Closed. The failing step now holds under the same conditions, and nothing that held before has started to fail.
- Not closed. Same step, same layer. The fix did not reach the path, or it addressed the wrong layer.
- Moved. The old step holds and a new failure appears higher up or somewhere else. That is progress and a new finding at once.
A retest that could not reproduce the original conditions says so plainly, and its pass is filed as a fresh result rather than a closure. That only works if the first run was built to be repeated: same harness, same parameters, saved and versioned, which is the property testing DDoS protection properly asks for from the very first run.