fix(swarm): accumulate execution_time across interrupt/resume cycles

[pgrayy]

Motivation

When a Swarm is interrupted and resumed, two time-related calculations are incorrect:

1. execution_time in SwarmResult is reset on resume instead of accumulating across invocations
2. elapsed time in should_continue only considers the current invocation, causing timeout checks to ignore time spent in previous invocations

The correct behavior is demonstrated in Graph, where both calculations properly account for accumulated time.

Resolves #1501

Public API Changes

No public API changes. These are behavioral bug fixes that ensure:

• SwarmResult.execution_time correctly reflects total execution time across interrupt/resume cycles
• Timeout checks in should_continue account for total elapsed time

# Before: execution_time was reset on each invocation
result = swarm("task")  # execution_time = 50ms (interrupted)
result = swarm(responses)  # execution_time = 60ms (only resume time, lost 50ms)

# After: execution_time accumulates
result = swarm("task")  # execution_time = 50ms (interrupted)
result = swarm(responses)  # execution_time >= 50ms (accumulated total)

Changes

File	Line	Change
swarm.py	202	elapsed = self.execution_time / 1000 + time.time() - self.start_time (matches graph.py)
swarm.py	409	self.state.execution_time += instead of = (matches graph.py)

Use Cases

• Performance monitoring: Total execution time across multi-step workflows is now accurate
• Timeout enforcement: Timeouts now correctly trigger based on cumulative time, not just current invocation time
• Cost estimation: Applications using execution time for billing receive correct totals

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[pgrayy]

/strands can you also look at the elapsed time calculation in should_continue?

[pgrayy]

/strands can you also look at the elapsed time calculation in should_continue?