Roadmap

Maintainance:

Release on the 1st Wednesday of each month
- markd: does train leave the station model still make sense for Toil?

More efficient kubernetes support
- One of:
  - Host-path caching
  - Toil-integrated within-pod scheduler
More robust Kubernetes support
- Handle Kubernetes communication timeouts without restarting the leader
Rework SingleMachineBarchSystem to eliminate thread limit exhaustion issues
CWL 1.1+ Support
- 13 failing CWL 1.1 conformance tests currently
- Conditional support (CWL 1.2)

Move away from mesos before/as Ubuntu 16.04 goes out of support
- probably in favor of auto-deployed Kubernetes somehow
Running in-house VG WDL workflows
WDL compliance test suite
Increase test coverage
Automatic idle worker termination and fixes to ignored nodes.
Updates on caching (should we enable by default?).
Incorporate a cactus integration test to better support cactus
Improved ease of debugging

Google Support
Batch System Support
Update boto libraries to boto3
Move from simpleDB to a better supported service
More scalable Kubernetes support * moving to watches * handling more pods in queue than we can loop over before our continue tokens expire
Restart/recovery improvements * Changing CWL parameters * Managing a failed task that cannot be recovered in a large pipeline * Checkpointing
Better support for heterogeneous tasks (e.g. customizing disk size per instance type, (maybe) FPGA support for DRAGEN).
AWS custom/multi security group support
AWS multi-zone support