catch stalls

noticed that the scan command stalls sometimes
conditions:
- many events from the same tx being processed by step_arb
- maybe others

current resolution:
- restart, or
- reduce batch_size and restart

Start with the scan command and the step_arb processing path, then reproduce the stall using many events from the same transaction. Done means scans continue without requiring a restart or reducing batch_size; the issue does not name a file or test, so locating the relevant code will require repository exploration.

Written by the indexing model from the issue text.