All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=759589

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=6d061b51bd100c91555d5b9afcdf623988b2c772151f4b91a79eff702a7843a9


Bot(s) for this bug's original alert(s):

chromium-rel-win10

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8970007121932674576

=== Auto-CCing suspected CL author pwnall@chromium.org ===

Hi pwnall@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Victor Costan
  Commit : c008fc280e48b6f15aeae4c552e695717a6b2d91
  Date   : Wed Aug 23 10:14:38 2017
  Subject: IndexedDB: Disable large value wrapping.

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : blink_perf.owp_storage
  Metric       : IDBFactory::open/IDBFactory::open
  Change       : 27.44% | 1.50532142626 -> 1.91831666629

Revision             Result                   N
chromium@496636      1.50532 +- 0.508347      14      good
chromium@496646      1.59239 +- 0.437559      9       good
chromium@496647      1.49682 +- 0.120689      6       good
chromium@496648      1.92146 +- 0.140525      6       bad       <--
chromium@496649      2.14702 +- 0.782892      9       bad
chromium@496651      2.03394 +- 1.22428       14      bad
chromium@496656      1.96006 +- 0.595701      9       bad
chromium@496675      1.91832 +- 1.13923       9       bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.owp_storage

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970007121932674576


For feedback, file a bug with component Speed>Bisection

cc: cmumford, dmurph, jsbell as they're also familiar with the situation

The CL disables a feature (large value wrapping in IndexedIDB) that was intended to bring a performance improvement. Sadly, as described in the bug attached to the CL, the feature interacts in an undesired (and unexpected) way with the blob system, causing us to "forget" about files we write on the user's disk. 

Thanks for the explanation pwnall! I'm marking this WontFix; the right people are cc-ed on the bug to reopen if I'm mistaken.