# Kill Or Cancel Does Not Stop Work

## Question

Was cancellation delivered, checked by the active execution path, and able to
release its tracked work?

## Enter When

Use when a query/task continues after kill, cancel, timeout, connection close,
or resource-control action.

## First Batch

1. Establish the cancel/kill request time, target identity, and whether the
   request was acknowledged.
2. Identify the execution phase/thread/task still active and its last
   observable progress or wait.
3. Check one boundary: signal propagation, cancellation check gap, memory/work
   tracking gap, supported-function restriction, or hung/buggy execution.

## Decision

Do not infer a product defect merely because work remained visible after the
request; distinguish delayed cleanup, uninterruptible phase, wrong target,
unsupported operation, and missing propagation. Validate version-specific
behavior before citing known fixes.

## Strongest Alternative And Stop

Check target-identity mismatch or delayed observability first. Stop after the
delivery/check boundary is located or one exact missing execution artifact is
named. Never issue kill/cancel automatically.