Categories
Extreme Programming Fit for Purpose Kanban Programming

Escape the Big-Rewrite Trap

If you’re a Staff Engineer parachuted into a degraded ecosystem with years of “quick fixes,” systems that don’t talk, recurring incidents, flaky pipelines. The instinct is to propose a big rewrite or a “new architecture initiative.” That’s structural change in disguise: it threatens ownership, timelines, and identity (“my service,” “my module”), and it triggers the technical version of a crisis. Productivity drops, defects spike, and leadership runs out of patience right at the bottom of the J-curve. The problem isn’t that modern architecture is bad. The problem is trying to buy long-term quality with a short-term shock the org can’t tolerate.

The lesson from evolutionary change is brutally practical: start with normative changes. The small, reversible engineering moves that reduce risk without forcing a social reset. Think “capability upgrades,” not “reorgs.” Close feedback loops first: a green CI as a non-negotiable, test seams around hotspots, and observability that turns incidents into data. Make integration sane: explicit contracts (OpenAPI/AsyncAPI/schema), versioning policies, consumer-driven contract tests, and anti-corruption layers that stop legacy weirdness from infecting new code. These moves don’t demand a new org chart, but they quietly dismantle the chaos tax.

Action is where a Staff Engineer earns their keep. Run a forensic baseline: map dependencies, then identify hotspots using churn + defects + incident history (the 10% of files causing 80% of pain). Create a “stabilization runway” with concrete guardrails: pipeline <10 minutes, “no broken builds after stand-up,” quarantine flaky tests, and ship smaller PRs daily. Then apply a strangler approach: pick one critical flow, carve out a clean use-case boundary, put ports/adapters around DB and external calls, and migrate slice-by-slice behind feature flags. Pair this with flow policies: WIP limits for urgent bugs, explicit Definition of Done (tests + telemetry), and aging WIP reviews to keep work from rotting on the board.

The strategic punchline: evolutionary technical change creates emergent structural change without forcing it. Once contracts exist and the build is trustworthy, teams naturally decouple. Once hotspots shrink, ownership becomes clearer. Once MTTR and change-fail rate drop, leadership stops demanding heroics and starts funding modernization. Track the ROI in executive-friendly terms: lead time, change-fail rate, MTTR, incident volume by hotspot, and % capacity spent on firefighting. You’re not “doing Kanban” or “doing architecture.” You’re installing evolutionary DNA so the system and the org around it can keep improving without another failed transformation.