ref(batcher): Only flush the bucket that triggered the flush event

[sentrivana]

Description

The span batcher uses multiple buffers (buckets) to store spans, keyed by trace_id. If a bucket fills up, we flush the whole buffer (all the buckets) instead of just the one bucket that actually should be flushed. This leads to too-frequent, busy flushes and is also against the spec.

With this change, we keep a set of buckets to flush. We still support full flushes for cases when sentry_sdk.flush() is called manually, or when the SDK is shutting down. Time-based flushes are also still global, i.e., not per bucket.

Issues

• closes Only flush trace buckets that triggered a flush event #5585
• closes https://linear.app/getsentry/issue/PY-2123/only-flush-trace-buckets-that-triggered-a-flush-event

Reminders

• Please add tests to validate your changes, and lint your code using tox -e linters.
• Add GH Issue ID & Linear ID (if applicable)
• PR title should use conventional commit style (feat:, fix:, ref:, meta:)
• For external contributors: CONTRIBUTING.md, Sentry SDK development docs, Discord community

[linear-code]

PY-2123 Only flush trace buckets that triggered a flush event

[github-actions]

Codecov Results 📊

✅ 13 passed | Total: 13 | Pass Rate: 100% | Execution Time: 8.04s

All tests are passing successfully.

✅ Patch coverage is 85.71%. Project has 14946 uncovered lines.

Files with missing lines (1)

File	Patch %	Lines
_span_batcher.py	82.93%	⚠️ 21 Missing and 11 partials

---

Generated by Codecov Action

[sentrivana]

This one was split in two (test_default_attributes and test_transport_format) and the latter was moved to test_span_buffer

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2b37852. Configure here.}

[alexander-alderman-webb]

LGTM!
It's probably not great having so many sleeps in the test, but the alternative is relying on private details (which wouldn't be good either).

[sentrivana]

It's probably not great having so many sleeps in the test, but the alternative is relying on private details (which wouldn't be good either).

Yeah definitely. Tried to keep them as low as possible at least.