Findings on the Continued Use of Deploy Friday Calculation
Field observations on the risk-assessment methodology that varies by hour and has no documented inputs.
Observation Log
Friday, 10:22 AM. Engineer evaluates a deploy. Feels good about it. The changeset is small. The tests are green. Confidence is described internally as "probably fine."
Friday, 12:04 PM. Engineer returns from lunch. Evaluates the same deploy. Confidence has shifted. The word "probably" is now doing more work. Deploy is rescheduled for Monday.
Monday, 9:48 AM. Deploy goes out. Everything is fine.
Three weeks later. Same engineer, same Friday, slightly different changeset. Process repeats.
Analysis
Under the governing rule, once deploy friday calculation passes locally, system-level doubt may be deferred to the next environment.
The engineer evaluates whether a Friday deploy is safe using confidence feelings that change sharply after lunch. This is not documented as a methodology. It functions as one.
The deploy friday calculation produces a local feeling of certainty while leaving system-level uncertainty untouched. The system appears calmer. The team records the symptom as resolved but not the mechanism.
The Pattern
Months later the same condition returns with enough history to make the temporary fix look like architecture.
The Monday deploy window, originally adopted as a one-time precaution, is now a team norm. Nobody remembers deciding it. Everyone follows it. New engineers are told this is how the team operates.
It is, in fact, how the team operates.
Status
The deploy friday calculation continues to function as described. No formal risk model has been proposed to replace it. The methodology is widely understood and will not be written down.