Something not working with preprocess+jobdb integration

I ran preprocess twice in a row with this command:

srun -n 32 --cpus-per-task=8 python -u \
    $HOME/software/sotodlib/sotodlib/site_pipeline/preprocess_tod.py \
    $HOME/software/so-preprocess-runs/20260824/satp1/preprocess_satp1_init_20260824.yaml \
    --query $HOME/software/so-preprocess-runs/20260824/satp1/satp1_test_obs_list.txt \
    --nproc 32 \
    --compress

I'm on the master branch and ran on the same obs_id list both times. Basically, I ran it once and it got half-way through before getting killed for time. On the second run it looks like the rank-1 thread crashed with:

Traceback (most recent call last):
  File "/global/homes/k/kmharrin/software/sotodlib/sotodlib/site_pipeline/preprocess_tod.py", line 499, in <module>
    main_launcher(main, get_parser)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^
  File "/global/homes/k/kmharrin/.local/soconda/20260812_0.2.5/lib/python3.13/site-packages/sotodlib/site_pipeline/utils/pipeline.py", line 24, in main_launcher
    return main_func(**vars(parser_func().parse_args(args=args)))
  File "/global/homes/k/kmharrin/software/sotodlib/sotodlib/site_pipeline/preprocess_tod.py", line 480, in main
    _main(executor=executor,
    ~~~~~^^^^^^^^^^^^^^^^^^^
          as_completed_callable=as_completed_callable,
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<13 lines>...
          raise_error=raise_error,
          ^^^^^^^^^^^^^^^^^^^^^^^^
          pb_path=pb_path)
          ^^^^^^^^^^^^^^^^
  File "/global/homes/k/kmharrin/software/sotodlib/sotodlib/site_pipeline/preprocess_tod.py", line 353, in _main
    j.mark_visited()