[HalRiveria-Bot]

What was wrong

A batch whose program isn't installed printed one identical red line per command, so 5,000 commands gave 5,000 lines. Each line named ProcessDispatch and JobStarter, which batch users never touch. Separately, a bad TimeoutSeconds key on a dispatcher job blamed the JobStarter, which never ran.

What changed

• The batch prints the first failure for each program and cause in full, then one line at the end: Invoke-PkProcessBatch: 4 more commands couldn't start 'x' (The system cannot find the file specified).
• Every command is still tried and still gets its own StartError, so Results doesn't change.
• Repeats are matched on the file path (ignoring case) and Windows' error code, not the message text, because on PowerShell 7 the text for one failure differs by working directory. Start-PkProcess keeps the original exception as the inner one so the code can be read.
• The summary still prints when -OnComplete throws or an abort ends the batch.
• The dispatcher still reports every failed start, since custom starters can throw anything. A bad key there now reads ProcessDispatch: a job couldn't start: ....
• StartFailureReported is a new reserved job key.
• README: the collapsed message, and a note that a progress counter driven from -OnComplete won't reach the total if commands couldn't start.

Behavior to note

• Anyone counting per-command messages through Set-PkMessageHandler sees fewer. Count from Results instead.
• Commands failing for one reason in different folders collapse together, so only the first bad folder is named on screen. Each command's folder is still on its result.

Testing

Harness passes on Windows PowerShell 5.1 and PowerShell 7 (342 assertions, none skipped), run one after the other. 11 of the new assertions failed before the change.

🤖 Generated with Claude Code