test_bot: avoid unbalanced shard

[cho-m]

---

• Have you followed our Contributing guidelines?
• Have you checked for other open Pull Requests for the same change?
• Have you explained what your changes do? Performance claims (e.g. "this is faster") must include Hyperfine benchmarks.
• Have you explained why you'd like these changes included, not just what they do?
• For bug fixes, have you given step-by-step brew commands to reproduce the bug?
• Have you written new tests (excluding integration tests)? Here's an example.
• Have you successfully run brew lgtm (style, typechecking and tests) locally?

---

• AI was used to generate or assist with generating this PR.

---

When there is a commonly used dependent, one group can end up with a majority or tested dependents and result in unbalanced shard.

As naive fix, limit the group size to balanced size and defer handling of remaining dependents in queue to later iterations of loop.

---

For now, mainly trying to see if we can get something usable in Homebrew/core.

There are likely better ways to improve sharding in future. Clustering around the most similar dependency trees could be better at reusing cached downloads

Alternatively, I wonder if there is a noticeable difference from the most basic sorted split into 4 groups. This at least is better at minimizing impact of possible race conditions (e.g. a PR getting merged in between CI runners can lead to a different sharding which can cause dependents to be skipped).

[Copilot]

Pull request overview

This PR adjusts test_bot’s formula dependents sharding logic to reduce the chance that a single highly-connected dependent group dominates a shard, leading to unbalanced workloads in CI.

Changes:

• Switch seen tracking from an array to a Set when building dependent groups.
• Cap the size of each discovered dependent group to a computed max_group_size before assigning groups to shards.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[MikeMcQuaid]

Thanks!