v0.6.2

What's Changed

Fixed

• Capital Efficiency section silently missing for operators on RPC nodes with receipt pruning — get_bond_event_history scans from CSM deployment block (20,873,000) in chunks; on nodes that prune transaction receipts (Nethermind's default, Geth in hybrid mode, etc.), the first chunks fail with -32000 "Receipt not available", the 3-strike abort fires, and bond_events returns []. apy.capital_efficiency is then dropped from the saved-operator JSON and the dashboard hides the section without explanation. The scan now distinguishes pruned-receipt errors (skip the chunk, don't count toward strikes) from real RPC failures (abort and surface a data_warnings entry + WARNING log). For operators created after the prune cutoff, all bond events are now collected even on pruned nodes.
• Empty bond_events result cached for a full hour — A scan that produced [] was held in the 1h @cached(ttl=3600) decorator, so even after fixing the underlying issue a stale empty result lingered. Caching is now inline with a conditional TTL: 3600s for non-empty results, 60s for empty results, so retries can recover quickly.

Changed

• Bond event scan is ~13x faster on the uncached path. Three changes combine: (1) the caller now passes a start_block derived from the operator's earliest distribution frame (with a 6-month safety margin) instead of scanning from CSM deployment; (2) chunk size raised from 10k to 50k blocks, consistent with other event scans; (3) the 8 bond event types are now scanned in parallel via asyncio.gather instead of sequentially. Measured against a local Nethermind RPC for an active operator (220 validators, 18 bond events), the uncached scan went from ~16 minutes to ~73 seconds. Cached results still hit in <1ms.
• Bond event scan abort now logs at WARNING (was DEBUG) so the failure is visible at the default INFO log level.

Upgrade notes

Users on pruning RPC nodes should restart the dashboard container after upgrading to clear the in-memory 1h cache that may be holding a stale empty result, then click Refresh on affected operators. First refresh takes ~1-2 minutes for the full bond-event scan; subsequent loads are instant.

Full Changelog: v0.6.1...v0.6.2