refactor(logs): one give-up budget for a --follow stream, not two

[davidsu]

Note

Description

base44 logs --follow tracked streaming failure in two independent counters: a connect-retry ladder (CONNECT_RETRY_DELAYS_MS, reset on every call) and a dead-stream budget (dropsSinceLastEvent, capped at 2 for the whole session). Since #595 made both give-up paths end the command identically, the split bought nothing and disagreed in both directions — the ladder could end a tail while the drop budget still held a spare life, and a backend alternating the two failure kinds could keep the CLI reconnecting forever because each counter was reset by the other's failure. This collapses them into a single strikes budget, where a strike is any attempt that produced no logs.

Related Issue

Follow-up to #595 (feat(logs): true realtime --follow via apper SSE stream). No linked issue.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Other (please describe): behaviour-affecting refactor — reconnect thresholds and pauses change, see "Additional Notes"

Changes Made

• Replaced dropsSinceLastEvent with a single strikes counter shared by both failure kinds (failed connect, and stream that opened then delivered nothing).
• RECONNECT_DELAYS_MS ([1s, 2s, 4s, 8s]) now doubles as both the backoff ladder and the give-up limit via hasRunOutOfStrikes().
• connectWhileTransientlyUnavailable → connectWithinStrikes(filters, strikes): continues the caller's budget instead of starting a fresh one, and returns the strikes it spent so streamUntilExhausted carries them forward.
• New Reconnected result type (stream | refused | exhausted) so an exhausted budget is distinguishable from a refusing backend.
• strikesAfter(ending, strikes): a server-side rollover (event: end, retriable: true) wipes the budget to zero; a stream that proved itself alive (a row or a keepalive) costs exactly one strike rather than earning a clean slate, so a flapping backend can't spin indefinitely.
• The post-stream reconnect pause is now the ladder rung for the current strike count instead of a flat STREAM_RECONNECT_DELAY_MS.
• Deleted MAX_DROPS_SINCE_LAST_EVENT, STREAM_RECONNECT_DELAY_MS, countDropTowardGivingUp, shouldGiveUpStreaming, connectWhileTransientlyUnavailable, and the now-unused LogStreamAttempt import.
• Rewrote tests/cli/logs.spec.ts's "stream drop budget" suite as "stream give-up budget" against the new helpers, adding cases for a budget shared across failure kinds and for rollover wiping it.

Testing

• I have tested these changes locally
• I have added/updated tests as needed
• All tests pass (npm test)

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation (if applicable)
• My changes generate no new warnings
• I have updated docs/ (AGENTS.md) if I made architectural changes

Additional Notes

Observable behaviour changes, all inside the --follow reconnect loop:

• The dead-stream threshold moves from 2 to the shared budget of 4 attempts, since drops no longer get their own small budget. This mainly helps quiet apps, where a stream can legitimately send nothing for its first ~20s (apper's KEEPALIVE_SECONDS) and a multi-pod rolling deploy could otherwise burn the cap of 2.
• The initial connect is now at most 4 attempts over ~7s of backoff (was 5 attempts over ~15s) before falling back to polling.
• The pause before reconnecting after a stream ends now scales with the strike count (still 1s when the budget is clean, as before).

Naming intentionally reads as prose (hasRunOutOfStrikes, ONE_STRIKE_FOR_A_STREAM_THAT_WORKED) to keep the retry policy legible at the call site. No file under docs/ describes this internal retry policy, so none needed updating.

⚠️ The streaming reconnect path still has no end-to-end coverage — pre-existing, and unchanged by this PR. Coverage here is unit-level on the pure helpers (strikesAfter / hasRunOutOfStrikes).

---

_{🤖 Generated by Claude | 2026-08-31 10:55 UTC | 90f0255}

[github-actions]

🚀 Package Preview Available!

---

Install this PR's preview build with npm:

npm i @base44-preview/cli@0.1.12-pr.611.23d31b2

Prefer not to change any import paths? Install using npm alias so your code still imports base44:

npm i "base44@npm:@base44-preview/cli@0.1.12-pr.611.23d31b2"

Or add it to your package.json dependencies:

{
  "dependencies": {
    "base44": "npm:@base44-preview/cli@0.1.12-pr.611.23d31b2"
  }
}

• 📦 Preview Package: @base44-preview/cli@0.1.12-pr.611.23d31b2
• 🔗 View this commit on GitHub

---

_{Preview published to npm registry — try new features instantly!}