Feature Flags, System Monitoring, and Developer Experience — Questions¶

This file tracks big-picture, cross-module questions about observability, feature flags, and Developer Experience. For questions specific to a single module, use that module's QUESTIONS.md instead.

How to Use This File¶

1. Write questions immediately — when something is confusing, log it here before moving on.
2. Be specific — "I don't understand SLOs" is less useful than "Why does my SLO need to be stricter than my SLA?"
3. Update status as you find answers — partial answers count.
4. Link your answers to the module or resource where you found the explanation.
5. Add follow-up questions — good answers often raise better questions.

Status Legend¶

Question Format¶

### Q{{N}}: {{QUESTION_TITLE}}

**Asked:** {{YYYY-MM-DD}}
**Status:** 🔴 Unanswered

**Full question:**
Write the question in full. Include context about what prompted it.

**My current hypothesis:**
What do you think the answer might be, even if you're not sure?

**Answer:**
_To be filled in_

**Source:**
_Where did you find/confirm the answer?_

**Follow-up questions:**
- _Any new questions this answer raised_

Questions¶

Q001: When should I use Prometheus vs. DataDog vs. OpenTelemetry metrics?¶

Asked: 2026-06-09 Status: 🟡 Partial

Full question: The topic covers Prometheus, DataDog, and OpenTelemetry — all of which can collect metrics. When would I choose one over another? Are they complementary or competing?

My current hypothesis: Prometheus seems best for infrastructure/self-hosted metrics; DataDog seems better for teams that want a fully managed solution; OpenTelemetry seems like a vendor-neutral abstraction layer. But I'm not sure how these interact in practice.

Answer: Research in progress. Check Module 02 (Prometheus) and Module 07 (DataDog) for detailed comparisons.

Source: To be filled in

Follow-up questions: - Can I use OpenTelemetry to send metrics to both Prometheus and DataDog simultaneously? - Is there a cost tradeoff between self-hosted Prometheus and managed DataDog at scale?

Q002: What is the difference between a feature flag and a configuration value?¶

Asked: 2026-06-09 Status: 🔴 Unanswered

Full question: Both feature flags and configuration values are runtime conditionals. What makes something a "feature flag" specifically? When does a config value become a flag, and when should I use a dedicated flag system like PostHog or LaunchDarkly instead of a config file?

My current hypothesis: Feature flags are configuration values specifically designed for controlling feature visibility, with additional capabilities like user targeting, gradual rollouts, and audit logs that plain config files don't provide.

Answer: To be filled in

Source: To be filled in

Follow-up questions: - What are the failure modes of using a database for feature flags vs. a dedicated service? - How should stale flags be cleaned up to avoid technical debt?

Q003: How do the three pillars (metrics, logs, traces) complement each other in practice?¶

Asked: 2026-06-09 Status: 🔴 Unanswered

Full question: The theory says metrics tell you something is wrong, logs tell you what happened, and traces tell you where the problem is in a distributed system. But in practice, when am I reaching for each one? Can you walk through a realistic debugging scenario that uses all three?

My current hypothesis: I think you'd start with a metric alert (latency is high), look at logs to find error messages, and then use traces to identify which service in a call chain is slow. But I haven't verified this workflow.

Answer: To be filled in — see Module 01 (Introduction) for the initial framework and Module 05 (Distributed Tracing) for the integrated debugging workflow.

Source: To be filled in

Follow-up questions: - Is there a name for this "metrics → logs → traces" debugging workflow? - How does correlation ID help bridge logs and traces?

Add new questions above this line. Keep them numbered sequentially.

Resolved Questions Archive¶

Move fully answered questions (🟢) here to keep the active list manageable.

(none yet)

Question Stats¶