Subscription exhaustion race causes deployment key cost unrecorded in baseline failover

[rohith500]

Problem

When a Claude subscription is healthy at the pre-dispatch exhaustion check (line 1794) but becomes exhausted during the primary OSS model dispatch (concurrent request burns the plan window), the baseline failover path has a race that causes the deployment key cost to be absorbed silently with no billing record.

Execution trace of the race:

Line	Action	ctx suppressed?
1794	claudeSubscriptionExhausted = false → suppression NOT applied to ctx	No
2051	Primary OSS dispatch fails; concurrent request flips snapshot to exhausted	No
2097	claudeSubscriptionExhausted = true — only baselineCtx suppressed, not ctx	ctx: No
2109	Baseline dispatch succeeds on Weave's deployment key	—
2147	ctx = resolveAndInjectCredentials(ctx, ProviderAnthropic) → subscription token injected	No ← bug
2200	servedOnSubscription(ctx) → true → delta = 0 → deployment key cost unrecorded	—

Three attributes are misattributed:

• cost.subscription_served OTel attribute — reports true instead of false
• SubscriptionServed billing field — causes delta = 0 instead of full cost debit
• credentialKeyPrefix/Suffix/credSource telemetry — attributes the turn to the subscription token instead of the deployment key

Fix

Add ctx = withSuppressedClaudeSubscription(ctx) alongside baselineCtx at the baseline exhaustion check (line 2097) so the re-resolution at line 2147 falls through to the deployment key correctly.

Fix available in PR #533.