I also tried my tool (Toma) with several problems
but all job pairs failed with the status
stage not reached (23).
By closer inspection into one of the benchmark problems
I found the following suspicious message:
this is the default benchmark type for rejected benchmarks and
benchmarks that are not associated with a type n=no_type.
I attached some screenshots for information.
After further inspection it seems that presence of a benchmark
processor does not matter in my case.
In my view, the problem is simply summarized as follows:
the status "stage not reached (23)" is shown when timeout is appropriate.
If this is the case, I think this could be easily reproduced
by setting a very short time-out.