fix(sampling): Attribute backpressure as unsampling reason more accurately

[sentrivana]

Description

Context

If backpressure handling is active, it lowers the effective sample rate.

There are different outcomes for when a span is unsampled due to sample_rate or backpressure.

Behavior

Before this change, if under backpressure, we'd only report backpressure as the reason for unsampling if the monitor lowered the effective sample rate to a falsy value. Which only covers a small portion of unsample decisions made due to backpressure. The rest would be reported as a generic sample_rate drop, potentially leading to confusing reporting. If someone e.g. has a traces_sample_rate of 1.0, they might see spans being dropped because of sample_rate, which is hard to justify.

In this PR, we try to do a better job at determining whether backpressure was actually the reason a span was dropped and emit the corresponding outcome.

The idea is, for an unsampled span:

• if sample_rand would've been higher than the original, non-backpressure sample_rate, the span would've been unsampled regardless of backpressure handling, so the outcome should be sample_rate
• if sample_rand actually would've been lower than the original sample_rate had the sample_rate not been lowered, the span would've been sampled. In that case backpressure handling is the direct cause and the outcome should be backpressure

Issues

Reminders

• Please add tests to validate your changes, and lint your code using uv run ruff.
• 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

[github-actions]

Codecov Results 📊

✅ 67 passed | Total: 67 | Pass Rate: 100% | Execution Time: 7.51s

📊 Comparison with Base Branch

Metric	Change
Total Tests	—
Passed Tests	—
Failed Tests	—
Skipped Tests	—

✨ No test changes detected

All tests are passing successfully.

❌ Patch coverage is 16.67%. Project has 14860 uncovered lines.
✅ Project coverage is 35.12%. Comparing base (base) to head (head).

Files with missing lines (1)

File	Patch %	Lines
sentry_sdk/tracing_utils.py	16.67%	⚠️ 5 Missing

Coverage diff

@@            Coverage Diff             @@
##          main       #PR       +/-##
==========================================
+ Coverage    35.12%    35.12%        —%
==========================================
  Files          190       190         —
  Lines        22901     22905        +4
  Branches      7854      7856        +2
==========================================
+ Hits          8044      8045        +1
- Misses       14857     14860        +3
- Partials       715       715         —

---

Generated by Codecov Action

[alexander-alderman-webb]

Cool code!