ai/live: Don't suspend orchestrator if the client has not sent data.

[j0sh]

Also retry the current segment on any other type of error

Another alternative to #3629 and #3614

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 23 lines in your changes missing coverage. Please review.

Project coverage is 30.54831%. Comparing base (b28a901) to head (19f85ef).
Report is 3 commits behind head on master.

Files with missing lines	Patch %	Lines
server/ai_live_video.go	0.00000%	23 Missing ⚠️

Additional details and impacted files

@@                 Coverage Diff                 @@
##              master       #3642         +/-   ##
===================================================
- Coverage   30.56062%   30.54831%   -0.01231%     
===================================================
  Files            156         156                 
  Lines          47162       47181         +19     
===================================================
  Hits           14413       14413                 
- Misses         31908       31927         +19     
  Partials         841         841                 

Files with missing lines	Coverage Δ
server/ai_process.go	1.68350% <ø> (ø)
server/ai_live_video.go	0.00000% <0.00000%> (ø)

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b28a901...19f85ef. Read the comment docs.

Files with missing lines	Coverage Δ
server/ai_process.go	1.68350% <ø> (ø)
server/ai_live_video.go	0.00000% <0.00000%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[leszko]

LGTM.

@j0sh wdyt about merging this PR, but also #3629 ?

[j0sh]

@leszko

wdyt about merging this PR, but also #3629 ?

I don't think we really need both, one or the other will do the trick, but this one gives us a little more grace because the rate of non-terminal disconnects is higher when excluding e2e tests ... the actual rate seems to be a few percent, rather than less than 1 percent with e2e included, but its hard to say exactly how much since we can only query so much in Grafana / Loki. See this note

FWIW I do prefer #3629 since it's less invasive, but I am cautious about impacting the overall experience too much, especially with the renewed focus on mobile where we're more likely to see intermittent "connectivity issues" due to backgrounding.

[j0sh]

Added a couple more fixes:

Better handle an outSegmentTimeout of zero

Don't suspend orchs on cancelled contexts

[leszko]

Added a couple more fixes:

Better handle an outSegmentTimeout of zero

Don't suspend orchs on cancelled contexts

Ok, sounds good. Thanks Josh!