I'm fixing up some broken benchmarks and changed one of them (that's data-driven, so has several test-cases) to use an internal time-out of 290s, comfortably less than the Watchdog's default 300 seconds.
After further fixes finally make each of the two test-cases that were previously failing (before the test had begun) now actually run, I find that each of them if run on its own times out as intended after 290s; however, when I run the two of them back-to-back, the second gets mauled by the watchdog timer five minutes after the first started - which is, of course, mere seconds after its own start.
I conclude that the watchdog timer is somehow failing to reset its timer between tests.

To reproduce: fetch this commit of qtbase, build the benchmark it modifies and run

qtbase/tests/benchmarks/network/socket/qlocalsocket/tst_bench_qlocalsocket 'pingPong:connections: 1000' 'pingPong:connections: 5000'

Try with either on its own, or reverse the order of the test-cases on the command-line.
Either on its own passes, but whichever gets run second second gets bitten by the other's watchdog.