BenchmarkInSitu segfaulting in HIP build of VTK-m

I have been trying to work on the BenchmarkInSitu benchmark in the HIP build of VTK-m. The InSitu benchmark appears to run to completion (almost), but I have observed that it consistently segfaults at the end during clean-up. It is not sporadic; it is repeatable and fails in the same way each time. A working hypothesis is that destructors associated with the input data sets are being called multiple times and leading to these segfaults.

@bolstadm mentioned that this problem also existed in the BenchmarkFilters benchmark, and has been corrected in the following commit: 6222baa7.

The InSitu benchmark is quite different compared to the filters benchmark, so translating the fix in 6222baa7 to work in the InSitu driver doesn't appear as straightforward as I initially thought (at least at this time). I would appreciate some targeted guidance on what needs to be done, or perhaps someone more familiar with VTK-m's internal design can help with this? I appreciate any help.

In case it is relevant, I have attached my build script that I used to build VTK-m. The error can be reproduced by simply running the BenchmarkInSitu executable. build_vtkm.sh

I have also attached output from the benchmark for reference: insitu.log