Prometheus metric spikes during KrakenJS plugin hot-reloading

When utilizing the @krakenjs/plugin-prometheus plugin to expose application metrics, there is an uncertainty regarding the lifecycle of metric counters during plugin refreshes. While the framework supports a plugin-based architecture, the precise timing of metric de-registration during a hot-reload is not clearly defined. n

The goal is to ensure that downstream monitoring systems do not receive transient spikes or duplicate entries when a plugin is reloaded. Currently, it is unclear if the Prometheus registry clears existing gauges or if they persist until the new instance is fully initialized.

How does the KrakenJS core handle the de-registration of metrics when a plugin hot-reload occurs? Is there a documented configuration to prevent transient metric persistence from triggering false alerts in the scraper during this transition?