feat(kiloclaw): support Northflank tier resize

[pandemicsyn]

Summary

• Adds Northflank support for existing-instance tier upgrades by growing the Northflank volume when needed and patching the deployment service compute plan before persisting the accepted desired tier state.
• Extends the provider adapter with an optional resizeRuntime hook so provider-specific resize semantics stay inside the provider layer while Fly and docker-local keep their existing behavior.
• Updates the admin resize flow to skip Fly-style stop/start orchestration for Northflank and show rollout-specific copy.

Verification

deployed regular instance , upgraded to 4x8

Visual Changes

Before	After
Admin resize dialog used Fly-style stop/start copy for every provider. Screenshot pending.	Admin resize dialog shows Northflank rollout-specific copy and storage growth warning for Northflank instances. Screenshot pending.

Reviewer Notes

• Main risk area is Northflank API semantics: volume update returns an empty success response with no separate completion status, so the Worker treats HTTP 200 as accepted and persists desired tier state after Northflank also accepts the compute-plan patch.
• Northflank resize intentionally does not require the instance to be stopped; Fly and docker-local resize behavior is unchanged.
• Billing remains tier-unaware. This only updates runtime tier state and the denormalized instance_type read cache.

[kilo-code-bot]

Code Review Summary

Status: 1 Issue Found | Recommendation: Address before merge

Overview

Severity	Count
CRITICAL	0
WARNING	1
SUGGESTION	0

Issue Details (click to expand)

WARNING

File	Line	Issue
services/kiloclaw/src/providers/northflank/index.ts	875	Resize is persisted before the Northflank rollout is confirmed

Files Reviewed (13 files)

• .specs/kiloclaw-controller.md - 0 issues
• apps/web/src/app/admin/components/KiloclawInstances/KiloclawInstanceDetail.tsx - 0 issues
• services/kiloclaw/src/durable-objects/kiloclaw-instance.test.ts - 0 issues
• services/kiloclaw/src/durable-objects/kiloclaw-instance/index.ts - 0 issues
• services/kiloclaw/src/northflank/client.test.ts - 0 issues
• services/kiloclaw/src/northflank/client.ts - 0 issues
• services/kiloclaw/src/northflank/config.test.ts - 0 issues
• services/kiloclaw/src/northflank/config.ts - 0 issues
• services/kiloclaw/src/providers/northflank/index.test.ts - 0 issues
• services/kiloclaw/src/providers/northflank/index.ts - 1 issue
• services/kiloclaw/src/providers/types.ts - 0 issues
• services/kiloclaw/wrangler.jsonc - 0 issues

Fix these issues in Kilo Cloud

---

_{Reviewed by gpt-5.5-20260423 · 1,220,687 tokens}

[kilo-code-bot]

WARNING: Resize is persisted before the Northflank rollout is confirmed

resizeMachine() persists the new tier immediately after this patch resolves, but patchDeploymentService only confirms the update was accepted. If the deployment later fails, the DO still records the larger instanceType, machineSize, and volumeSizeGb, and the admin UI reports completion even though runtime hardware may not have changed. This path should wait for the deployment to complete (as startRuntime/restartRuntime do) before returning success or persisting tier state.

[pandemicsyn]

I don’t think we should wait here. The persisted tier is the desired state after provider acceptance, its also consistent with the Fly resize path, which persists instanceType/machineSize before the restarted Fly machine proves it is running on the new CPU/RAM.

If Northflank accepts the patch and later fails rollout, that is runtime/provider health drift for inspect/reconcile to surface, not a reason to keep the desired tier old.

[jeanduplessis]

Looks good overall. I agree with the existing Northflank resize thread as a non-blocking improvement suggestion: waiting for rollout completion before marking the resize done would reduce provider/state drift when Northflank accepts a patch but the rollout later fails. I did not add a duplicate line comment because that thread already covers the point.