This is becoming a real pain. Developers are receiving messages with "Failing for the last 1042 test runs ", failed since build #0 which is completely false.
As explained above, this happens after a test has been Skipped.
If the Skipped build is removed/deleted everything is back to normal again and the correct "Failing for the last 5 test runs" is displayed again.

When a build is skipped all Age results are set to "1"
( see attached pictures - build#513 was also skipped )
Pictures show another example with builds 512/514/515

Further analysis:
Failed build shows <failedSince>642</failedSince>
Skipped build : <failedSince>0</failedSince>
all subsequent builds : <failedSince>0</failedSince>

The fix is:
For a skipped build, update the failed since to the previous value, do not set to zero.

junit plugin - Worked in version 1.27 - fails in 1.28
probably the fix for the stack overflow.

https://plugins.jenkins.io/junit/
Changelog
Version 1.28 (May 14 2019)
This plugin now requires Jenkins 2.138.4 or newer.

JENKINS-31660 - Do not load attempt to load test history for skipped tests. This fixes some cases where a StackOverflowError was being thrown.

See:
https://github.com/jenkinsci/junit-plugin/pull/117
There are two methods that recompute the failedSince field, one of them handled skipped builds correctly, the other one did not. With this PR they will both use the same logic.

Found your name from the list of changes to jUnit.
I reported this defect - it should be clear in the description of what is happening.
The failedSince is needed for the skipped and subsequent builds.

Thanks for the investigation, I'll try to fix it soon.

had to upgrade to version 1.29, only noticed since a test failed for the last 598 days when it was skipped in the last build and passed in the build previous to that.
I went back to 1.27 and then screwed up all my warnings generations since there is a dependency in many other plugins for 1.29..
Reloaded 1.29 but my warnings generation is still messed up - there are now no reported warnings.
Hows it going with a fix?

Any progress with this ?
I have test results which have been skipped for 1346 test runs ( job number is #1345 ) - its not correct.
I have hundreds of developers really making fun of the results and I have to continually explain what happens.
Thanks..

Still broken in LTS 2.277.1

sgjenkins this was released in JUnit 1.50 today

The fix won't correct existing data, if a test is marked as failing since build 1, new builds will still show that it's failing since build 1. But after it's skipped once the age counter will reset.