Status report: 2026-10-17 14:16 UTC

Original request: speed up TinySite page loads (with parallel prefetch), let TinySite take about 8 uploads at once, add the two management-page lists (unused files and recent 404s, with three-part timestamps), race GitHub-runner builds against the remote build service so slow remote builds cannot block merging, fold in the TinySite large-file effort (land its three pull requests, deploy them, prove a 35 MB file serves whole), and end by proving filedrop is working, since filedrop was touched throughout. Added 08:40: the owner delegated the five open decisions to me (constraint: never raise a container's keep-warm period beyond 5 minutes). Added 10:25: the owner reported https://tinysite.wasmserver.com/ showing an error. Added 13:09: the owner asked that merged changes be deployed as appropriate after each merge.

🟡 Status: RUNNING. Every remaining step is mine; no question is open.

Progress since the last report (14:07 UTC)

• All six lanes launched at 14:05 are alive and writing (every output file touched within the last minute). Early results from their findings:
  • Filedrop pin-bump review (https://github.com/CodexCoder21Organization/FiledropServiceServer/pull/34) has one medium finding so far: the eight-upload test's bound on shared bookkeeping writes is unsound, because the fake storage only counts the commit call and not the payload writes, so a request can be waiting on a healthy payload write while the test reads it as "queued its manifest update". It observed 8 manifest and 16 listing writes for eight uploads through the real RPC path, against the author's claim of at most 2 and 6. The verdict is pending; the PR's own checks are all green (only the informational remote-build check is still running).
  • Externalbrain round 3 did not fix the remote failures. The required remote check on https://github.com/CodexCoder21Organization/ExternalBrainWui/pull/58 at 0024696 fails the same two tests as before: the recovery scenario's HTTP read times out on the shared build service, and the permanent-failure scenario sees a null where it expects the second attempt. Locally both pass in about 40 s (the round-3 reviewer just ran them). So the remote failure is not the real-time deadline the author converted; the mechanism is still open. I saved the remote outputs and will brief the next author round with them, together with my own finding (the request router constructs a fresh document-page servlet per request, which re-reads the page template every time) and the reviewer's verdict when it lands. The reviewer has already pushed one passing edge test (immediate failure with concurrent requests) and is writing a combined shutdown case.
  • Read-idle regression fix is pushed (https://github.com/CodexCoder21Organization/UrlProtocol/pull/721 now at d1e8af9): the lane confirmed both reviewer tests fail 3/3 on the old head first; branch checks are running on the new head.
  • kompile-cli upstream fix has its mechanism written before any code: the manager library's close owner inspects the completed stop failure synchronously twice (for terminal cancellation, then while building its warning), and a lease-renewal error is formatted while renewal is still marked in progress; tests will inject failures whose message and cause getters block on a test-controlled future at each boundary. The upstream clone (about 572 MB) is still transferring.
  • kompile-cli CI diagnosis has the retry-fixture test passing repeatedly on main in a separate worktree and is now running the new SIGTERM stress fixture there, up to ten times, to establish whether the two exit timeouts are pre-existing.
  • Buildtest consumer validation refuted its first hypothesis: the two startup-cleanup scenarios pass on clean main with the published journal 0.0.13, so they are not pre-existing; the candidate changes documented startup-observation contracts that those tests still assert in the old form. It is now reproducing the mid-page compaction failure.
• The consumer PR's CI is worse than my last report said: https://github.com/CodexCoder21Organization/BuildTestEmbedded/pull/1490 has eight failing tests across six of the sixteen shards (the four I briefed plus an undrained-restart admission test, a 120-second timeout in a paged legacy-journal test, and two more classification-file startup tests), and bld-build is red only as the aggregate of those shards. The running lane was briefed with four; the full list is saved for its follow-up round if its fix does not cover the rest.
• UrlProtocol coordinate bump https://github.com/CodexCoder21Organization/UrlProtocol/pull/720 is still first in its merge queue; the queue's full-suite run has been in progress since 13:31 (the previous queue run in this repository took 41 minutes).
• TinySite was not re-probed this cycle; its first visit after idle still depends on the filedrop redeploy.

Pull requests

✅ Merged this effort: https://github.com/CodexCoder21Organization/TinySiteWui/pull/7 (12:16 UTC), https://github.com/CodexCoder21Organization/FiledropEmbedded/pull/25 (12:12), https://github.com/CodexCoder21Organization/UrlProtocol/pull/718 (12:00), https://github.com/CodexCoder21Organization/FiledropServiceServer/pull/33 (11:58), https://github.com/CodexCoder21Organization/FiledropEmbedded/pull/19 (10:41), https://github.com/CodexCoder21Organization/TinySiteWui/pull/6 (09:59), https://github.com/CodexCoder21Organization/SimpleFileSystemServiceEmbedded/pull/38 (08:53), https://github.com/CodexCoder21Organization/FiledropApi/pull/6 (08:48), https://github.com/CodexCoder21Organization/FiledropEmbedded/pull/22 (06:18), https://github.com/CodexCoder21Organization/TinySiteWui/pull/5 (05:58), https://github.com/CodexCoder21Organization/HandoffServiceServer/pull/33, https://github.com/CodexCoder21Organization/HandoffEmbedded/pull/28, https://github.com/CodexCoder21Organization/FiledropEmbedded/pull/18, https://github.com/CodexCoder21Organization/FiledropEmbedded/pull/20, https://github.com/CodexCoder21Organization/HandoffServiceServer/pull/32, https://github.com/CodexCoder21Organization/HandoffEmbedded/pull/29, https://github.com/CodexCoder21Organization/BuildTestProjectionJournal/pull/12, and four challenge records (https://github.com/CodexCoder21Organization/PlanRepository/pull/11034, https://github.com/CodexCoder21Organization/PlanRepository/pull/11033, https://github.com/CodexCoder21Organization/PlanRepository/pull/11036, https://github.com/CodexCoder21Organization/PlanRepository/pull/11040). Published: filedrop-embedded 0.1.27, filedrop-api 0.0.7, simplefilesystemservice-embedded 0.0.32, projection-journal candidates rc4 and rc5. Deployed: TinySite bbce49a (13:14 UTC); handoff service 0.0.30.

⏳ Pending (live at 14:13 UTC):

Delegated work at a glance

⏳ Running - GPT 6.1-sol high - 8 minutes total time - under 1 minute since last update - Independent review of the filedrop server pin bump https://github.com/CodexCoder21Organization/FiledropServiceServer/pull/34 at 2f8ff4b. So far: one medium finding that the eight-upload test's write-batching bound is unsound (the fake storage counts only commits, not payload writes) and an observation of 8 manifest and 16 listing writes through the real RPC path. Next expected checkpoint: the remaining rows (legacy-read proofs, rewritten fixtures, the connection-library coordinate side question) and the verdict.

⏳ Running - GPT 6.1-sol high - 8 minutes total time - under 1 minute since last update - Round-3 review of https://github.com/CodexCoder21Organization/ExternalBrainWui/pull/58 at 0024696: pushed a passing immediate-failure-with-concurrent-requests edge test to its review branch, ran the converted recovery scenarios locally (passing, about 40 s each), and is writing a combined shutdown-during-attempt test. Next expected checkpoint: its verdict.

⏳ Running - GPT 6.1-sol high - 8 minutes total time - 1 minute since last update - Fixing the regression in https://github.com/CodexCoder21Organization/UrlProtocol/pull/721: confirmed both reviewer tests fail 3/3 on the old head, pushed the fix as d1e8af9; branch checks running. Next expected checkpoint: the 73-file selection green, PR description updated and marked ready.

⏳ Running - GPT 6.1-sol high - 7 minutes total time - 1 minute since last update - Closing the round-11 finding on https://github.com/CodexCoder21Organization/kompile-cli/pull/192: mechanism written (two synchronous inspections of the stop failure by the close owner, plus a lease-renewal error formatted mid-renewal); reviewer reproducer running in forked children against manager 0.0.207; upstream clone transferring. Next expected checkpoint: upstream fail-first tests committed.

⏳ Running - GPT 6.1-sol high - 7 minutes total time - under 1 minute since last update - Diagnosing the three bld-all-tests failures on https://github.com/CodexCoder21Organization/kompile-cli/pull/192: the retry-fixture test passes repeatedly on a main worktree; now running the new SIGTERM stress fixture there up to ten times with a bounded controller. Next expected checkpoint: pre-existing-or-introduced classification for each failure.

⏳ Running - GPT 6.1-sol high - 7 minutes total time - under 1 minute since last update - Consumer validation of https://github.com/CodexCoder21Organization/BuildTestEmbedded/pull/1490: refuted "pre-existing on main" for the cleanup scenarios (2/2 green on clean main with journal 0.0.13), identified that the candidate's documented startup-observation contract changed under tests still asserting the old callbacks; reproducing the mid-page compaction baseline failure now. Next expected checkpoint: classification of each of its four failures.

⚪ Not running (finished/dead) - GPT 6.1-sol high - 2 minutes total time - 41 minutes since last update - TinySite route configuration: refuted the premise with source evidence (the site cache deletes its run directory on stop and clears abandoned ones at start) and made no production change. Nothing further is expected from it.

⚪ Not running (finished/dead) - GPT 6.1-sol high - 1 hour 35 minutes total time - 36 minutes since last update - Buildtest round 4: published rc5 (byte-verified), journal PR 109/109; consumer full run reached its cap without a result. Replaced by the consumer-validation lane above. Nothing further is expected from it.

How the fixes work

• TinySite management lists, prefetch cache and large-file streaming (merged, deployed 13:14): every request updates an in-memory record written off the request path with per-entry generations; a bounded on-disk cache is filled in parallel by owned storage calls that stop interrupts and drains, one version per answer; files are served in 524,286-byte ranges from the published API constant, and the uncached path confirms size and upload time after every range.
• Filedrop old-drop read, manifest read and group commit (merged, published; server pin bump in review): a serving read returns the drop's original complete metadata in two storage calls regardless of file count; one ranged read of the whole manifest; uploads touching the same bookkeeping document are batched so one leader does one read and one write per batch with no lock held during I/O.
• Filedrop API constants (merged, published) and the server-side fix (merged): the download range and legacy limits are compile-time constants; the served client validates the caller's whole range before splitting and treats a padded part as terminal.
• Storage library (merged, published): the manager's close interrupts and joins its trash-reclaim workers before releasing the root and refuses to join from inside a callback that holds a manager lock.
• Connection library (merged, coordinate bump in the queue) and its read-idle follow-up (fix pushed): a completed answer that does not fit the response queue waits instead of being dropped; each paused answer has one owner-tracked deadline retired exactly once; separately, pending handlers count as activity so the idle check cannot close a stream under them, and bookkeeping for a permit released after the loop stopped runs inline or not at all.
• kompile-cli (upstream fix in progress): the race's remote run is stopped on wait expiry or exit with a one-time fallback; every exit passes through one completion boundary that cannot itself fail; no code on the shutdown path, local or in the manager library's close, reads a failure's message before the outcome is published.
• Buildtest CPU (consumer validation in progress): journal compaction does per-row bounded work and streams its staging delta; the startup row cache is bounded by the decoded footprint it retains.
• Handoff service (deployed): report bodies are archived out of the live state file so the heap no longer grows per report.
• ExternalBrainWui startup (round 3 in review; remote still red): the HTTP server binds before any dependency is contacted; dependencies open lazily with at most three attempts, each cut at 30 seconds on the injected clock; each route is gated only on the dependency it uses.

Blockers

🟡 None; every item has a lane running or a brief ready for the next slot.

Landing parallelization sweep: all landing gates are running in parallel: the filedrop pin bump has CI green, its review running and my final review done, so it enqueues the moment the review returns CLEAN; the externalbrain PR's review runs alongside its (red) remote check and the next author brief is pre-staged with the remote outputs; the read-idle fix's CI started on push. Nothing dispatchable is idle; the six-lane cap is full.

Plan changes: the externalbrain remote failure is reopened as an unexplained mechanism (the real-time-deadline theory is refuted by the run at 0024696) and goes into the next author round with my servlet finding; the buildtest consumer repair will need a follow-up round for the four failures outside the current brief unless its fix covers them.

Next steps: land the filedrop pin bump and redeploy filedrop, merge and publish the UrlProtocol bump then bump and redeploy the storage server, fix and land the read-idle, kompile-cli, externalbrain and buildtest PRs as their rounds come back clean and deploy what applies, make the TinySite cache survive restarts, then prove an 8-file parallel publish, old and new sites loading cold, faster page loads and a working filedrop.

Handoff

url://handoff/handoffs/hf-2026-10-17-land-the-tinysite-speed-and-parallel-upload-fixes-finish-the-ci-capacity-fixes-and-redeploy-tinysite-filedrop-and-storage, viewable at https://www.handoff.wasmserver.com/handoffs/hf-2026-10-17-land-the-tinysite-speed-and-parallel-upload-fixes-finish-the-ci-capacity-fixes-and-redeploy-tinysite-filedrop-and-storage; this report also reached the folded-in large-file handoff, whose remaining item is the filedrop server redeploy. Lane artifacts: https://github.com/CodexCoder21Organization/PlanRepository/tree/wip/tinysite-ci-handoff-artifacts-2026-10-17.

Top lessons

1. A fix for a remote-only failure is unproven until the remote run on the fixed head is read. The externalbrain deadline conversion was a sound theory (real 30-second waits on a slow shared host), passed locally, and still failed the same two tests remotely, so the theory was wrong or incomplete. Treat "passes locally" as necessary, not sufficient, for a failure that only shows on another host, and read that host's output for the fixed head before briefing the next step on the old theory.
2. Read the lifecycle code before treating a persistent path as persistent data. A configuration-only "make the cache survive restarts" task was refuted in minutes once the delegate read the cache's start and stop methods: the application deletes its own cache directory on stop by design. Before proposing an environment or path change to preserve state, confirm from the owning code and its tests that the state is meant to outlive the process.
3. When a review names a defect in a dependency, fix it there and pin the result, even mid-round. The kompile-cli exit blocker survived ten local rounds because its last instance lives in the pinned manager library's close path; patching around it locally would have hidden it from every other consumer. Route the fix to the owning repository with a fail-first test, publish a candidate, and bump the pin.