feat(profiles): env-gated discovery + event-driven dropdown health dots

[8144225309]

Background

The post-#46 orchestrator audit surfaced three related issues with the wallet's profile system:

1. The dropdown's per-profile red/green health dots are almost always stale — the existing probe loop only fires while nodeInfo.error is truthy (i.e., when you're already broken).
2. Auto-discovery on auth pollutes the cwd-relative node-profiles.json by adding every local CLN node it can find, breaking isolation for the demo (8 profiles instead of the intended 2 after a single restart).
3. checkAllHealth probes all profiles in parallel — fine at small counts, but a connection burst with many profiles.

This PR addresses all three, opt-in by env var where appropriate.

Changes

A — Event-driven dropdown health dots

No continuous polling. Per the design discussion, dots refresh only on:

• Profile switch — NodePicker.handleSwitchNode fires NodesService.healthCheck() in the post-switch background refresh.
• Rescan for Nodes button — NodePicker.handleDiscover fires it after the rescan completes.

The slice's setProfileHealth now merges by profileId instead of replacing the whole array. Partial probes (e.g. one capped by WALLET_HEALTH_PROBE_MAX) don't clear previously-known colors, and a profile that went red stays red until the next good probe answers from that profileId — the sticky-red behavior you asked for. Entries are never auto-removed.

B — Env-gated discovery + standardized profile-file path

Two new env vars:

Var	Default	Effect
WALLET_AUTODISCOVER	(unset / true)	Set to false to skip the auth-time local-socket scan entirely — the wallet uses only the profiles in the file you handed it
WALLET_PROFILES_FILE	./node-profiles.json (cwd-relative, backward compat)	Absolute path override; lets isolated/test runs point at a deterministic fixture instead of relying on cwd

Production: both unset → current friendly auto-discovery, current file location.
Isolated demo / orchestrator: both set → frozen, deterministic profile list every restart.

D — Sequential, capped probe in checkAllHealth

Promise.allSettled(profiles.map(probe)) → sequential for…await loop with two new env vars:

Var	Default	Effect
WALLET_HEALTH_PROBE_DELAY_MS	200	Delay between probes
WALLET_HEALTH_PROBE_MAX	64	Hard cap on profiles probed per scan; excess gets a logger.warn

One commando connection at a time. Worst case at 64 profiles, 200ms delay, 5s probe timeout ≈ a sequential scan that takes a few seconds — acceptable since it only fires on switch / explicit refresh.

Validation

• CI build/eslint/test on this PR.
• Post-deploy on the demo: set WALLET_AUTODISCOVER=false + WALLET_PROFILES_FILE=/root/ss-walletdemo/node-profiles.json in env.sh, restart, confirm dropdown returns to just the 2 demo profiles. Switch between them and watch the dots flip.
• Playwright orchestrator pass (sequenced after deploy) to repeat the cross-profile click-through that exposed R8.36: rendezvousSlice reducers + selectors tests #157 and verify the new sticky-red + event-driven dot semantics.

Files changed

File	Why
apps/backend/source/service/node-profiles.service.ts	WALLET_PROFILES_FILE env override
apps/backend/source/service/node-manager.service.ts	WALLET_AUTODISCOVER=false early-out + sequential capped checkAllHealth
apps/frontend/src/components/ui/NodePicker/NodePicker.tsx	healthCheck calls in both handlers + setProfileHealth import
apps/frontend/src/store/nodesSlice.tsx	merging setProfileHealth reducer