feat: persistent node should be able to start in ready mode waiting for POST /config

[cbaugus]

Problem

There are currently only two node startup modes:

Mode	Env var	Starts as	After test
Persistent	EPHEMERAL=false (default)	running — starts workers immediately, requires TARGET_URL	standby
Ephemeral	EPHEMERAL=true	ready — waits for POST /config	idle (one-shot)

There is no mode for a long-running node that starts idle and waits for a test to be pushed via API, then returns to ready after the test completes so it can accept another.

Desired behaviour

A persistent node should be startable with no initial test config:

1. Node starts in ready state — no workers, no TARGET_URL required
2. Operator pushes a test via POST /config
3. Test runs for its duration
4. Node returns to ready state (not idle, not standby)
5. Operator can push another POST /config → repeat

This is the natural model for a pool of always-on load test nodes managed by a control plane.

Proposed implementation

Add a NODE_MODE env var (or reuse EPHEMERAL semantics):

NODE_MODE=ready   # starts ready, returns to ready after each test (persistent)
NODE_MODE=ephemeral  # existing EPHEMERAL=true behaviour (one-shot)
# unset = existing default (starts running immediately from env vars)

Or simpler: when EPHEMERAL=false and TARGET_URL is not set, start in ready mode instead of erroring.

Current workaround

EPHEMERAL=true without SELF_DESTRUCT_CMD — node goes to idle after test but still accepts a new POST /config. Works but semantically incorrect (reports as ephemeral, node_state is idle not ready between tests).

Related

• bug: workers keep running past TEST_DURATION — time_remaining_secs goes negative, standby spawns wrong worker count #109 — workers keep running past TEST_DURATION
• bug: POST /config does not update workers or RPS on a long-running persistent node #110 — POST /config not updating workers/RPS on persistent node