Only a small percentage of tests can expose bugs, and extreme cases are often dismissed as unrealistic.

• Testing reveals the limitations of our imagination in finding bugs.

00:01

The video discusses a bug in the code that caused call records to remain on the stack even after the caller hung up.

• In the early 1980s, customer service centers had a high rate of callers being taken off hold within two minutes.

• The code didn't handle the case when the caller hung up, resulting in call records remaining on the stack.

• Originally, the code checked the stack every time something was added or removed, but it was removed later due to system time constraints.

• To prevent stack problems, a stack reset command was added in places where the stack was supposed to be empty.

02:23

Whole forwarding is a process where held calls are transferred to other phones, causing a rotating outage in the system.

• Whole forwarding transfers held calls to other phones.

• In a rotating outage, error handling triggers failures in subsystems one after another.

• The outage clears when call pace slows down, but during busy times it can last for hours.

• The bug causing the rotating outage could have prevented potential customers from getting help under life-critical circumstances.

04:46

Long sequence bugs are difficult to find and are not easily detected with coverage or error-oriented testing.

• Long sequence bugs only manifest after a long series of events.

• Testers are expected to perform various tasks in addition to testing, but the time required for testing is infinite, leaving little time for other activities.

• The decision to fully document tests is complex because it takes time away from other important tasks.

• Statements that testers should always do certain tasks are unreasonable because every task is subject to trade-offs.

• Hiring more testers does not solve the problem of an infinitely large set of testing tasks.

07:10