Create a parameterized job, append naginator plugin, set fix retry time to e.g. 2 minutes and retry count to 3. Then and add a short action which can fail the build depending on the parametervalue.
1) Launch the job with a parametervalue which causes the job to fail.
2) Wait until naginator has started the job again (1. retry)
3) Wait until naginator has started the job again (2. retry)
4) during the waiting time for the third retry, start the job with parametervalue for a successful build.
5) Wait until naginator has started the first (failing) job again (3. retry)
The first job which fails all the time will not be started again after the third retry.
Seems the successful build in (4) was resetting the count of (3) because the failed job is rebuild two times again.