[Tracking] Follow-up issues and baseline findings related to PR #65126

## Current status after connector SPI merge (#64304)

Updated after #64304 was merged into master and #65126 was rebased on top of it.

| Tracking item | Current status |
|---|---|
| #65750 HMS incremental-event consistency | Partially fixed: item 1 is fixed by #65126; items 2 and 4 were structurally fixed/removed by #64304; items 3, 5, and 6 remain open. A new item 7 now tracks missing connector-cache invalidation for database/table CREATE, DROP, and RENAME events. |
| #65779 names snapshot rebuild cost | Open; immutable NameCacheValue single-item updates still rebuild O(N) state. |
| #65792 unified cache statistics | Open and expanded: direct FE catalog/database entries and connector-owned entries are both absent from the unified statistics path. |
| #66001 global invalidation ordering | Open; #64304 does not provide one ordering barrier across FE names/object/ID state. |
| #66025 delegated-session SHOW TABLES N+1 | Open; the current getTables path still repeats remote name enumeration. |
| #66024 MetaInfoAction privilege filtering | Open; unrelated to #64304 and #65126 cache behavior. |
| #66233 ambiguous local-name mappings | Fixed by the current #65126 NameCacheValue implementation, pending PR merge; not fixed by #64304. |

### Current PR-only review items not closed by #64304

- Legacy ID-only refresh replay can still lose invalidation when the database object entry is cold.
- An unrelated key in the same MetaCacheEntry stripe can suppress required ID-map publication; the active-key side-effect fence is still pending.
- Targeted DROP/rename scans the retained ID map under the cache publication window; this remains a performance follow-up.

### Scope decision for connector event invalidation

The missing connector invalidation for HMS database/table events is a correctness gap introduced by #64304's cache-ownership split, but it is already present on the new master baseline. It is tracked in #65750 and does not have to block #65126 as long as the PR claims only FE metadata-cache convergence and the follow-up risk is explicit.

## Summary

This issue aggregates the follow-up issues, review findings, and baseline problems tracked outside PR #65126:

- PR: #65126
- PR URL: https://github.com/apache/doris/pull/65126

The goal is to provide one place where reviewers and later contributors can quickly find the problems that were discovered during the review of PR #65126, including both:

1. issues introduced or clarified during the refactor follow-up process, and
2. pre-existing master/baseline problems that were surfaced by the same review thread but intentionally tracked out of scope.

## Tracked issues

### Correctness / consistency

- #65750 `[Bug] HMS incremental events have multiple metadata cache consistency gaps`
- umbrella issue for HMS incremental-event consistency problems
- includes the mode-1 database-event canonicalization gap identified during PR review

- #66001 `[Bug][Meta Cache] Complete global invalidation ordering after PR #65126`
- tracks the remaining global invalidation ordering/fence gap after the per-key publication ordering fixes

### Performance / behavior follow-ups

- #65779 `[Performance] Reduce full names snapshot rebuild cost on HMS events`
- tracks snapshot rebuild overhead on HMS event paths

- #65792 `[Enhancement] Include catalog and database MetaCacheEntry instances in unified cache statistics`
- tracks follow-up observability work for the refactored cache entries

- #66025 `[Performance] Avoid N+1 remote table enumeration in external session bypass SHOW TABLES path`
- baseline performance issue surfaced during the PR review
- not introduced by PR #65126, but related to the same external metadata/session-bypass area

### Other related findings found during review

- #66024 `[bug](http) MetaInfoAction.getAllDatabases() leaks unfiltered db list, bypassing SHOW privilege`
- separate pre-existing master bug surfaced while reviewing the full PR scope

## Notes

- This issue is intentionally a tracking index only.
- Individual bug discussions, fixes, priority, and closure should happen in the linked issues.
- New issues discovered while reviewing or following up PR #65126 can be appended here later.

Start with the linked follow-up issues #65750, #65779, #65792, #66001, #66024, and #66025, then review PR #65126 and the baseline merge #64304 for context. This issue is an index rather than a standalone task; done means resolving or updating the individual findings and keeping their status accurately reflected here.

Written by the indexing model from the issue text.