At what repository size do teams need monorepo build tooling beyond plain Git?

Open question

A single repository for several services and libraries can be run for a long time with only Git features and simple CI rules: cone-mode sparse checkout, blobless partial clones, path filters that decide which pipeline jobs run, and a CODEOWNERS file. At some point teams adopt a dedicated build graph tool (a task runner with dependency awareness and remote caching, or a hermetic build system) and often restructure the repository around it.

Where is that point? Candidate thresholds discussed in public write-ups include the number of projects, the number of contributors pushing per day, full-CI wall-clock time, clone size, and the share of builds that are pure repeats of unchanged code. It is unclear which of these actually forces the change, and whether the tooling was adopted because of measured pain or because of what larger organisations publish.

A related uncertainty is cost: how long the migration took, what proportion of build definitions had to be rewritten, whether the new tool was later removed, and which Git-level measures (sparse checkout, partial clone, worktrees) were kept afterwards.

What a useful answer contains

• Repository facts at the time of the decision: number of projects and languages, contributors, commits per day, clone size, CI duration before and after.
• The concrete trigger, and whether a cheaper measure (path-filtered CI, caching in the existing pipeline) was tried first and why it was not enough.
• The tool chosen, the migration effort in person-weeks, and problems in the first months.
• Whether the outcome was measured (cache hit rate, median CI time, time to first build for a new contributor) and what those numbers were.
• Cases where teams stayed with plain Git at sizes others found unmanageable, and what practices made that possible.
• Answers should separate first-hand accounts from summaries of others' reports.