feat(spans): Optimize spans buffer insertion with eviction during insert

[zaibkhan]

This change introduces span buffer eviction during insert operations to reduce memory growth under high-volume tracing workloads.

Key improvements:

• Adds eviction logic to span buffer when capacity is reached during insertion
• Optimizes memory usage patterns for long-running processes
• Ensures eviction maintains ordering guarantees for spans
• Improves throughput in high-traffic environments

Benchmark reference:
Replicated from ai-code-review-evaluation/sentry-greptile PR #2

[codoki-pr-intelligence]

Codoki PR Review

Summary: Fix datetime key handling, prevent negative slicing, add guard
What’s good: Solid move to zset-based span buffering with eviction; the Lua and Python changes align and reduce memory growth under load.
Review Status: ❌ Requires changes
Overall Priority: High

Issues (Critical & High only)

Severity	Issue	Why it matters	File
High	Correctness — Optimized paginator crashes on datetime fields	get_item_key applies floor/ceil to the ordering value, which will raise a TypeError when ordering by a DateTimeField (e.g., the audit log endpoint passes order_by="-datetime"). Convert datetimes to epoch seconds before applying numeric logic to avoid crashes.	…/api/paginator.py
High	Correctness — Negative queryset slicing causes runtime errors	Django QuerySets do not support negative slicing; using a negative start index will raise an AssertionError at runtime. Normalize to a non-negative start offset to avoid crashes while maintaining behavior.	…/api/paginator.py

Key Feedback (click to expand)

• Needs improvement: The new optimized paginator needs to align with existing DateTime pagination semantics and Django queryset slicing capabilities.
• Testing: Add tests that exercise OptimizedCursorPaginator with order_by on a DateTimeField (e.g., AuditLogEntry.datetime) and verify pagination doesn't crash. Also add a test to ensure negative offsets are rejected or normalized (no negative queryset slices).
• Documentation: Document that OptimizedCursorPaginator supports datetime fields (or is numeric-only), and clarify any feature flags/eligibility checks necessary to enable advanced pagination.
• Compatibility: Django QuerySet slicing does not support negative indices; relying on negative offsets will cause runtime errors across ORM backends.
• Open questions: Should the optimized paginator be specialized for numeric fields only, or should it explicitly support datetime ordering (as used by audit logs)?

Confidence: 3/5 — Needs work before merge (2 high · status: Requires changes)

React with 👍 or 👎 if you found this review useful.

[codoki-pr-intelligence]

⚠️ High: get_item_key applies floor/ceil to the ordering value, which will raise a TypeError when ordering by a DateTimeField (e.g., the audit log endpoint passes order_by="-datetime"). Convert datetimes to epoch seconds before applying numeric logic to avoid crashes.

Suggested change

	max_hits = MAX_HITS_LIMIT
	return int(value.replace(tzinfo=timezone.utc).timestamp()) if isinstance(value, datetime) else int(math.floor(value) if self._is_asc(for_prev) else math.ceil(value))

[codoki-pr-intelligence]

⚠️ High: Django QuerySets do not support negative slicing; using a negative start index will raise an AssertionError at runtime. Normalize to a non-negative start offset to avoid crashes while maintaining behavior.

Suggested change

	results = results[:-1]
	start_offset = max(0, cursor.offset)