non-urgent SoD follow-up issues

[hjoliver]

Placeholder for possible changes to the initial SoD implementation as discussed in the SoD proposal, e.g.:

• [ ] try to avoid all remaining iteration over task proxies
• [ ] can we make prerequisite updates more efficient? (current use of pre-SoD methods may be OK though)
• [ ] extend SoD to clock- and external-triggers? (tasks could be spawned in response to trigger events)
• [ ] do we need stuck prerequisite housekeeping?
• [ ] better failed task handling?