Practical guides for platform engineers — Kafka, APIs, rate limiting, latency.
Why average load misleads, the 70% NIC rule, replication and consumer fanout math, and how partition counts — not headline throughput — often drive broker count.
How NFA backtracking explodes, nested quantifiers and overlapping alternation, real CVEs, safe engines and fixes, and why length limits plus testing beat hoping your regex is fine.
Why bumping max_connections makes things worse, how PgBouncer transaction mode works, the work_mem multiplier trap, and the prepared statement gotcha that breaks Rails and Django.
What DORA actually measures, why your overall tier equals your worst metric, how to avoid gaming the numbers, and how to get a baseline without SaaS tooling.
RecordBatch vs Record overhead, why linger.ms=0 kills compression, zstd vs gzip benchmarks, and the replica.fetch.max.bytes silent killer.
Fixed window boundary exploit, token bucket burst math, retry storm prevention, and why banking APIs need different configuration than public APIs.
What counts as a breaking change, why request and response schemas follow different rules, and how to version safely without breaking clients.
Updated Jeff Dean latency numbers for cloud-native 2026: NVMe, cross-AZ Kafka, Lambda cold starts, and why N+1 beats network as the top latency killer.
A practical guide to consumer lag math: partition bottleneck, max.poll.interval.ms misconfiguration, and the alerts you need before production incidents.