Bug Description

Assuming I'm not making any errors with my input, the develop branch seems to be having issues transporting on DAGMC geometries when a weight window mesh is used. Specifically, the ray firing routine has trouble finding a surface intersection, resulting in a lost particle with a warning like the following:

WARNING: No intersection found with DAGMC cell 1, filled with material 12

I suspect this is happening to new particles after splitting, but have not verified.

It's worth noting that in my testing, this issue occurs only when a weight window mesh is used. I cannot replicate with an analog or implicit capture run, even when bumping up the number of particles by several orders of magnitude.

I also cannot replicate the issue if using the latest release (0.15.2), with nothing else changed.

Steps to Reproduce

I've included an input stack for a minimum-working example. Note that this stack does not include weight windows (the resultant tarball is too large) but does include two separate scripts to create some. One script uses MAGIC and the other FW-CADIS, whichever is preferred. Note that if using the develop branch, the MAGIC method will fail after the first batch. For FW-CADIS, a small 6-group cross-section file is included.
spherical_shell.tar.gz

Also included is a DAGMC geometry of nested spherical shells, pictured below. The shells are filled with typical fusion materials. Note that the innermost sphere is implicit complement (void), illustrated in white. There is also a void gap region between the vacuum vessel and "magnets".

All material definitions correspond to the tags in the DAGMC geometry. With the exception of a weight window file for transport.py, one should be able to use the scripts as-is. Note that the scripts only create the model XML file. One would need to execute openmc after executing the script. The number of particles, batches, etc. may want some tweaking for testing, though. They're unnecessarily high.

Environment

Using the openmc/openmc:latest-dagmc and openmc/openmc:develop-dagmc Docker images.

Please let me know if any other relevant information is needed. Thanks!