parallel gem 1.18.0 causes higher CPU load and stall HTL

[zunda]

Expected behaviour

After merging the changes from tootsuite/mastodon, my server continues to work as before.

Actual behaviour

Merging the changes below, the container running Puma and Sidekiq started to show higher CPU load average. The home time line stopped showing new toots from remote or local.

• c31bc41 Bump @babel/core from 7.4.5 to 7.7.2
• ee820ca Bump @babel/plugin-proposal-decorators from 7.6.0 to 7.7.0
• acf132b Bump @babel/runtime from 7.7.1 to 7.7.2
• 4cade42 Bump enzyme-adapter-react-16 from 1.14.0 to 1.15.1
• 58faefe Bump babel-plugin-react-intl from 3.1.3 to 3.4.1
• a3a109d Bump parallel from 1.17.0 to 1.18.0
• b0c4eb2 Bump rubocop from 0.75.1 to 0.76.0
• 6fa2f3e Bump faker from 2.6.0 to 2.7.0
• 169ce49 Bump annotate from 3.0.2 to 3.0.3
• e311131 Bump stoplight from 2.1.3 to 2.2.0
• bc1af89 Bump react-notification from 6.8.4 to 6.8.5
• 6cbaaac Bump uuid from 3.3.2 to 3.3.3
• 11cb5e6 Bump @babel/plugin-proposal-class-properties from 7.5.5 to 7.7.0
• ac1709c Bump @clusterws/cws from 0.15.2 to 0.16.0
• c696093 Bump fuubar from 2.4.1 to 2.5.0

Steps to reproduce the problem

See above.

Specifications

Reverting a3a109d #12358 seems to have fixed the two issues.

[zunda]

Looking further the changes in parallel between 1.17.0 and 1.18.0 seem minimal and should not affect behavior of Puma or Sidekiq. I'll continue to dig further.

[zunda]

Tried the same code again and the problem didn't reproduce. Sorry for the noise!