Hardening Sprint

Myth: Every few Sprints, we need a Hardening Sprint to make the product robust
Variations:

• Test Sprint
• Bugfix Sprint
• Integration Sprint

Category: General Scrum myths
Danger: High

The basis of the myth

Many teams come from a project-oriented background where "after the development work is done", if any time is left, remaining effort is put into stabilizing the product. As Scrum is product-oriented and intended to "last forever", these teams frequently suffer from accumulated technical debt. A well known solution is the Hardening Sprint to stabilize the product so that it can be shipped.
This myth was supported by the old use of "Potentially shippable product increment" (PSPI) in the Scrum guide, which gave leeway to the interpretation "Well, we don't ship just yet...", which has since been abolished in favour of the "Increment".

Why is it a myth?

The Scrum Guide clearly states that an Increment must be Done. This is explained literally as: "This Increment is useable, so a Product Owner may choose to immediately release it." There is no leeway to accumulate technical debt until a later date.
The Scrum guide does provide a certain amount of leeway for teams who are not sufficiently adept at Software Craftsmanship or work with an unsupportive product to weaken their Definition of Done to an appropriate level. This implies that team members and the organization understand their constraints and are actively working to continuously improve their DoD.
The institutionalization of a "Hardening Sprint" is contrary to the spirit of agility, namely: releasing value early and often.

Assumption #1: We can do the work later
Later, there will always be more work. That's the idea of a Product Backlog. At any point in time, the Product Backog contains the most valuable things to do. Any work for a backlog item that is deferred to the "Hardening Sprint" is either sufficiently important to do now - or sufficiently worthless to do never.


Assumption #2: We'll fix the bugs later

Many teams institute hardening Sprints because they don't have test results yet. A common misconception is that another team does the QA for the last Sprint while the developers code more stuff. To ensure that the current Sprint does not get disturbed by bugfixing, the bugs are postponed until the Hardening Sprint, where they are consequently eliminated.
Well - there are many issues here:

• Scrum does not separate functions. The idea of a separate test team is un-Scrum. "Tests passed" would be part of a good DoD. 
• Development and test should always be in sync. Practices like BDD etc. help in designing a defect-free product from the beginning.
• As code always builds on preceding code, a defect found "down the line" will cause significantly more rework than a defect just created: Hardening Sprints create massive waste!

Consequences

Poor quality mindset

Developers should be proud of creating a product that is usable at any time. When they get used to the idea that "later, we'll get time to increase the quality", they will learn that low quality is acceptable.
As typically, there is always something more important and always one more bug, the product accumulates technical debt which is usually just kept at bay instead of systematically managed. 

Value denied

When the team relies on an upcoming "Hardening Sprint" before releasing an increment, the value of any increment is Zero until then. The incurred delay for delivery may be quite costly for the product organization.


Branching hell

As the product itself is "hardened" at a later date, emergency fixes are typically created in "branches". Since each fix needs to be stable both on Production and in the Increments, each fix will require an independent "bug branch". All of these then need to be merged during the Hardening, resulting in exactly the problems which Continuous Integration intends to fix.

Remedies

Abolish the concept of a "Hardening Sprint". It induces tremendous waste.

Think very carefully about which activities are considered necessary during a "Hardening Sprint". List them out explicitly and ask the following questions: "Why do we need this activity?" and "What would need to change so that we either don't need this activity any more?" If the latter is inconclusive, ask, "What would we need to change so that we can do this activity during each Sprint?"

Common solutions to fix the need for Hardening Sprints include:

• A strong Definition of Done
• Test Automation
• Firm Engineering Practices, such as: BDD, TDD, Continuous Integration
• Collective Code Ownership
• Truly Cross-Functional teams (including Design, Development, QA and Ops)
• Feature teams
• Giving Developers a Prod-like development environment (thanks @Docker!)