[ML] Influencer string with very large size may cause the autodetect process to crash

[valeriy42]

Customers have observed the behavior when the autodetect process would stop unexpectedly. A possible cause of the failure is an influencer value of >77K characters passed to the process.

The behavior was observed on v3_linux_system_information_discovery job.

We need to reproduce and investigate the possible causes for this failure.

[tankha]

Hello,

As temporary solution: as recommended by the support, we cloned the job with the runtime field, that cut the field to 16K chars, and this solved the problem