Inkpad

• Responded to reviews and support emails (pretty heavy week because of the ongoing release, still catching up)
• Discovered a bug in our Stripe upgrade flow for Android WebViews, but gave up when I realized it only affects really old versions of Chrome (e.g. version 33)
• Searched for the cause of a slow memory leak, but finally gave up and switched our App Engine instance type back to F4 for now.
  • Instances still get shut down periodically, but F4’s increased RAM will make it a less common occurrence. Less frequent shutdown means less double-billing for the overlapping 15 minute period when a full RAM instance is shutdown and a fresh one is started up.
  • Hopefully the migration to Python 3 will reveal/remove whichever library is the cause
• Disabled the more expensive dedicated memcache service (there was no meaningful difference in hit-rate compared to the cheaper “best-effort” memcache)
• Lots of app and server debugging, spelunking through logs.
• Removed App Engine’s appstats builtin. This was part of the memory leak search, but it needed to go anyway as part of the move to Python 3

Misc

• Reviewed our GCP billing logs - discovered an old virtual machine that’s been eating a few dollars/month and was intended for temporary use