All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=802401

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=7fb7945663be6031ceae210c928e2696b3617dadd1070bd4fb81b53d0b4b80d5


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

chromium-rel-win7-gpu-nvidia

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/148f97c0840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12ee0a48840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14c4b1f8840000

Pinpoint jobs don't seem to ever finish. Kicking off one more time before looping in dev team...

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/14c4b1f8840000

yanggou: Sorry the bisects are so hard to read, but if you look at this bisect:
https://pinpoint-dot-chromeperf.appspot.com/job/14c4b1f8840000

It does repro the drop at your CL, "Do not use soon deprecated v8::ReturnValue::Set." (r528947). Can you take a look?

This is very interesting. That CL merely replaces a deprecated API by its successor. I'll revert and verify whether that helps.

Then again maybe the benchmark does not track real-world usage? The name sounds a bit like a micro-benchmark...

Yep, this is definitely a microbenchmark. +jbroman, the owner, for guidance on how important this regression is to address.

This has been reverted here last week: https://chromium-review.googlesource.com/c/chromium/src/+/893358

The graph has indeed recovered.

Before I close this bug however, I'd like a verdict on how important the micro-benchmark is.

It's not super important itself (so if we're progressing other things, it'd be okay to regress slightly here). But this is somewhat a measure of the minimum cost of an API call into Blink, and it'd be nice to not slow them down if we can avoid it.