What is a Spike?
A spike is a timeboxed investigation whose deliverable is knowledge, not product: research a technology, prototype an approach, chase down a risk — then throw the code away and keep the answer. Technical spikes probe implementation questions; functional spikes probe user-facing ones.
The timebox is the discipline: a spike answers "can we, and roughly how?" — it is not a license to gold-plate a prototype into production.
Worked example
Before committing a sprint to real-time collaboration, the team runs a 3-day spike: can their stack hold 500 concurrent WebSocket editors? Day 3's answer: yes, but only with a Redis pub/sub layer — which changes the story's estimate from 5 points to 13 before the sprint bets on it. The prototype is deleted; the 8 points of foresight are kept.