fix: handle provider status default cases across providers.

[garvitssoni]

Problem

Across several providers, unknown/transient status values were being coerced into concrete lifecycle states (commonly pending). During deletion this led to incorrect “Starting/Deploying” signals and contributed to UI status regressions.

Root cause

Default-case handling in provider status mappers treated unknown values as known states instead of surfacing “unknown.”

Fix

• Standardize provider mappers to return “unknown” (empty/unset lifecycle) for unrecognized provider status values, rather than defaulting to pending/failed.

[stephahart]

Suggested change

	lifecycleStatus = ""
	lifecycleStatus = v1.LifecycleStatusEmpty

[stephahart]

How do we handle the empty status in dev-plane and the UI? In these cases, does it make more sense to assume the instance is terminated rather than returning an empty status?

[garvitssoni]

How do we handle the empty status in dev-plane and the UI? In these cases, does it make more sense to assume the instance is terminated rather than returning an empty status?

Hi @stephahart,

On the UI side, an empty provider status is handled as a separate case and displayed as “Deleting.” This differs from “Terminated,” which is shown as “Stopped.” Combining these would present users with an inaccurate view of the instance state.

On the dev-plane side, we track when an instance last entered Running, Stopped, or Terminated using the ProviderStatusTransitions struct. Cloud providers may occasionally return “not found” errors, which can resemble termination—even though the instance may not really be terminated yet. To ensure data accuracy, we only set LastTerminatedAt when termination is explicitly confirmed. If the status is empty or unknown, we do not record any transition time.

In practice, an unknown status indicates “not yet confirmed,” prompting frequent polling (every 15 seconds). A confirmed Terminated state is treated as final—we record the timestamp and reduce polling to every 60 seconds.

If we default unknown statuses to Terminated, we risk recording an incorrect termination time, slowing down polling prematurely, and potentially missing instances that recover later.

Thanks!