What Nigerian SRE interviews actually ask in 2026
Twenty-two hiring managers, eleven cities, one uncomfortable pattern about take-home exercises.
Systems design still dominates, but the framing has shifted from scale to cost. From the interview loops I have run and reviewed this year, nine of twenty-two panels now include an explicit budget constraint.
Take-homes over four hours correlated strongly with candidates dropping out. The teams I saw moving fastest replaced them with a paired debugging session.
The most common opening prompt in my notes this quarter was not "design a social network". It was "you own a service that already works, but your cloud spend doubled in ninety days; show us the first three things you would check". Candidates who started with request patterns, data-transfer paths and idle capacity usually did better than those who jumped into architecture diagrams.
Observability questions are also getting narrower and more practical. Several panels asked candidates to read a synthetic incident timeline and point out one missing metric that would have reduced time to recovery. The strongest answers were specific: queue depth on one path, saturation on one dependency, and an alert threshold linked to user impact rather than host CPU.
On reliability, runbooks are now part of hiring conversations. Four teams asked for a short runbook section in plain language, then asked what would break at 2:00 am when the original author is asleep. That caught many people who wrote perfectly structured documents with no validation steps and no rollback line.
The interview process itself is changing too. Teams that moved to a ninety minute paired debugging round reported better signal on collaboration, and they reduced offer cycles from three weeks to eight days. Candidates said the format felt closer to real work, and acceptance rates improved.
One uncomfortable finding from my own debriefs: several candidates with strong production experience were filtered out by quizzes focused on trivia. The panels that changed this kept one fundamentals checkpoint, then weighted real failure analysis higher. Those teams closed senior roles faster without lowering bars.