Compatibility of `lazy` and `remove`

[samwaseda]

In the recent issue in atomistics, I raised the problem that the input is not properly recovered when the job is supposed to rerun. What is supposed to happen in original pyiron is here:

1. Load job -> all job data is transferred to the job instance
2. Delete HDF5
3. Run

However, in the case of SPHInX, and probably what’s gonna follow many codes, these steps do not work, because in the loading step the data is not loaded when lazy=True, and when the HDF5 is not there anymore, the job cannot find the data, and hence no rerun. The same problem happened with the master jobs some time ago, because ParallelMaster tries to re-initialize the job id with similar steps. I have the feeling that we need a systematic approach to make sure that this kind of problem doesn’t happen anymore. This being said, I don’t really have a nice idea right now XD

[samwaseda]

I quoted a wrong issue. Here’s the one that I wanted to: https://github.com/pyiron/pyiron_atomistics/issues/959