Syslog agent panics when a valid Cloud Foundry task name exceeds RFC5424 PROCID capacity

Summary

The loggr-syslog-agent panics when processing logs from a Cloud Foundry task
with a long task name.

The task name is accepted as valid metadata by Cloud Controller. However, the
name is propagated through Loggregator as part of source_type and is later
used to construct the RFC5424 PROCID field.

Observed behavior

The syslog agent crashes with:

panic: runtime error: slice bounds out of range [:-40]

The panic occurs in:

src/pkg/egress/syslog/tcp.go

The relevant code is:

sourceInstance = sourceInstance[:(128 - len(sourceType) - 3)]

Expected behavior

A valid Cloud Foundry task name must not be able to crash the syslog agent.

The syslog agent should:

• safely truncate the complete generated PROCID to the RFC5424 limit, or
• reject/drop only the affected message and continue processing.

The agent process must remain running.

Impact

A single task with a sufficiently long name can cause the syslog agent on a
Diego Cell to restart. This can interrupt or impair delivery to configured
syslog drains and may cause message loss during the restart.

Start in src/pkg/egress/syslog/tcp.go at the sourceInstance slicing shown in the report, then trace the surrounding PROCID construction. Exercise processing with a valid Cloud Foundry task name longer than the RFC5424 capacity. Done means the generated PROCID is handled safely and the agent continues running instead of panicking.