Quick answer: Bug bashes that test 'everything' produce shallow coverage. Bashes that focus on the past month's crash hotspots produce deep coverage of the riskiest code.
Bash focus should be informed by data. The data exists; use it.
Past month crash hotspots
Top 5 systems by crash count. Bash focuses 70% of time on these.
Per-system test plan
For each hot system, 10-20 specific test cases. Cover the known repro patterns.
Free-form 30%
30% time is unstructured exploration. Catches the unknown bugs.
Post-bash review
Did the bash find new bugs in the focused systems? If not, the focus was wrong; iterate.
“Data-driven bashing is more efficient. The data tells you where to look.”
Track bash effectiveness per quarter. Focus areas should yield bugs; if not, refocus.