BF-39058 found a place in the RecordId implementation where we construct a std::string from a const char * taken from a StringData without passing the StringData length field. This results in a buffer overrun when calculating the length based on a NULL terminator when the underlying buffer isn't NULL-terminated. Regardless of whether it's correct for a RecordId to have a non-NULL-terminated underlying string, we should adapt the code in RecordId to be more resilient to buffer overruns.

Either ensure that all std::string objects that are constructed in RecordId from StringData objects see the length field from the originating StringData or validate the NULL-termination of the StringData on ingestion (or both).