usage tracking pickle failure

[benclifford]

Describe the bug

In CI runs I am seeing this suspicious looking debug message sometimes:

 DEBUG    parsl.usage_tracking.usage:usage.py:199 Usage tracking end message (unencoded): {'correlator': 'f7161d2b-1b0f-4401-b705-b0d79e31588f', 'end': 1764589853, 'execution_time': 9, 'components': [{'c': 'parsl.dataflow.dflow.DataFlowKernel', 'app_count': 1, 'app_fails': 0}, {'c': 'parsl.config.Config', 'executors_len': 1, 'dependency_resolver': False}, {'c': 'parsl.executors.high_throughput.executor.HighThroughputExecutor', 'mpi': False}, 'parsl.providers.local.local.LocalProvider', 'parsl.launchers.launchers.SingleNodeLauncher', 'parsl.data_provider.ftp.FTPInTaskStaging', 'parsl.data_provider.http.HTTPInTaskStaging', 'parsl.data_provider.file_noop.NoOpFileStaging', 'parsl.monitoring.radios.filesystem.FilesystemRadio', 'parsl.monitoring.monitoring.MonitoringHub'], 'project_name': 'parsl htex_local_alternate test configuration'}
DEBUG    parsl.usage_tracking.usage:usage.py:212 Usage tracking failed: Can't pickle <function udp_messenger at 0x7fb86c266340>: it's not the same object as parsl.usage_tracking.usage.udp_messenger

I have not investigated what is going on here. It's reported as DEBUG not ERROR, but thats perhaps coming from a time where the sneaky "you won't notice our usage tracking" crowd was trying to hide breakages rather than make them visible.

If it is actually a failure, this should be tested and fail in CI.

To Reproduce

I haven't attempted to reproduce this

Expected behavior

Usage tracking should work. Usage tracking should be tested. Breakage in usage tracking should be reported as an ERROR and also as a CI test failure.

Environment
CI