Fetch uncached V3 pools on demand so cold quotes find liquidity

[AryanGodara]

Description

The driver's Uniswap V3 fetcher warm-caches the top max-pools-to-initialize pools ranked by raw liquidity. Raw liquidity isn't comparable across pools (it scales with token decimals and price range), so well-funded pools with a low raw value are excluded from the warm cache. When a quote requests such a pair, checkpoint.get() returns nothing and fetch returns zero AMMs, failing the quote until a later maintenance tick happens to back-fill the pool.

This change makes fetch fetch the requested pairs' uncached pools on demand instead of deferring to maintenance, so the first quote for any pair finds liquidity.

Changes

• PoolsCheckpointHandler::get also returns the ids of pools that exist for the requested pairs but aren't cached.
• fetch fetches those on demand (new fetch_pools) before building the result; on failure it serves whatever is cached.
• fetch_pools skips pools whose conversion fails (e.g. ticks not yet available) instead of aborting the batch; update_missing_pools reuses it, so periodic maintenance gains the same resilience.
• Unit tests for the on-demand path and the skip-on-failure behaviour.

Stacked on #4559 (bootstrap split) and #4555 (migration isolation) so one image carries all three for combined staging validation.

How to test

Unit + e2e tests.
Did a local driver against an ink-seeded pool-indexer. it loads the previously-missing pools on the first quote.

[gemini-code-assist]

Code Review

This pull request implements on-demand fetching for Uniswap V3 pools that are missing from the warm cache, allowing them to be fetched at the indexer's current block instead of waiting for the next maintenance cycle. However, two high-severity issues were identified: first, if no pools exist for the requested token pairs, the checkpoint block number defaults to 0, which triggers a massive event replay from block 1 and can cause database overload or OOM; second, passing 0 as the target block to fetch current pools breaks the subgraph client backend by querying the genesis block. Both issues require immediate fixes to ensure stability and compatibility.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.