Ryan Thorpe

For the time being, I've created a new file called `mpi_comm_spawn_child.py` that more-or-less mirrors the style of `mpi_child.py`. Eventually, these files should be combined into one. It also currently runs,...

Let's setup a time to chat more about this!

FYI, I think the reason `MPIBackend` is setup the way it is currently (i.e., prior to this PR) is specifically to provide timeouts + tracebacks when errors occur prior to...

> seems like this might be a way to go: https://mpi4py.readthedocs.io/en/stable/mpi4py.run.html using an `-m` flag? Try to avoid peppering the codebase with `try/except` blocks ... it's really anti-pattern. But if...

Minor update, I finally got this branch to run off of a master MPI process on Oscar (Brown's HPC). See [here](https://gist.github.com/rythorpe/31fddcdddf13e234bc3ba970a9f7628d) for a demo.

Isn't this issue more about consolidating the seed values set by the user in a few different places throughout the API into a single seed value?

I guess this will be really easy to resolve once we can completely get rid of legacy mode, since we'll then be able to set the seed once and only...

Ahh good point. Then we could just pass around a single `Generator` instance (i.e., that has been seeded in `simulate_dipole()`) so that we still address the original concern brought up...

Closing due to redundancy with #244. Feel free to re-open if there's is something unique here that should be addressed separately.

> OK, `drives.py:drive_event_times` is the remains of `feeds.py`, i.e., deprecated (used to be called `feed_event_times`). The test `test_drives.py:test_external_drive_times` is basically testing that the expected "default" drives are read from the...