How to force a “flush” / reset state between multiple input files in the same FME workflow?

Hello everyone,

I’m working on an FME workspace where I need to process multiple input files sequentially within the same workflow, and I want to make sure that each file is processed independently, without any data or state being kept in memory from the previous file.

Conceptually, what I’m looking for is a way to flush or reset the internal state between files, so that:

• cached features,
• temporary aggregates,
• statistics,
• or in-memory objects

from a previously processed file do not affect the processing of the next one.

Context

• I’m looping over multiple files (using a file-based workflow, not separate workspaces).
• The workflow includes transformers that may keep internal state (e.g. aggregations, statistics, grouping, geometry operations).
• The goal is to guarantee deterministic, per-file processing, similar to a “clean run” for each file.

My questions

1. Is there a native way in FME to explicitly flush / reset the internal state of a workspace between input files?
2. Are there specific transformers or workspace settings that:
   • automatically reset state per file, or
   • should be avoided when processing multiple files in the same run?
3. Is the recommended approach to:
   • use a workspace per file,
   • use feature-based partitioning (e.g. Group By),
   • or rely on specific transformers (e.g. FeatureHolder, FeatureMerger, etc.) to isolate state?