Ye Luo

> Maybe some crossed language here: once created, walkers have fixed ID independent of timestep. So if a run aborts due to population explosion (most common case), we just follow...

When using my index(parent index). Step 0. rank 0: walkers 0(0),2(2) rank 1: walkers 1(1),3(3) step 1. rank 0: walkers 0(0),2(2) rank 1: walkers 1(1),3(3),5(1) # both 1 and 5...

I'm thinking about a fast way of backtracing a bad walker and accessing infos like coordinates and energy on the evolution path in the analysis phase. This analysis is usually...

Here is the issue of immutable walker id. If walker id is based on the creation MPI rank not the actual rank were the walker locates, looking up the walker...

@PDoakORNL could you clarify "rank id's are local"? What is rank id you were referring? My understanding is that walker id is defined in each step and thus meaningful for...

Here is an example of the bubble I was referring to. When starting 3 walkers on rank 0, 1 and 2 walkers on rank 2, 3 for total request 10...

If the walker id is only intended to generate a unique id, then the scheme in this PR is ok. However, it cannot be used to be more informative indicating...

> I believe in practice walkers should not be getting moved more than necessary so the ID's should be presevered between steps unless the walker actually changed rank or unless...

> Walker ID is not mutable. I walker that migrates is killed and born again. Unfortunately killing a physical term not a technical term. Migranting a walker cannot be treated...

In the above [example](https://github.com/QMCPACK/qmcpack/pull/5010#issuecomment-2148617363) I was talking about gaps between ranks not within a rank. ids spaced by num_ranks were not considered disjoint.