Lambda /tmp directory not cleared before invocation reuse

[stormchasing]

When calling the lambda for hundreds of files at once I was getting failures due to lambda invocations that were running out of the available 640 MB. I raised the memory limit to 1024 MB and again saw out-of-memory failures. I eventually discovered this was due to files from previous lambda runs being left in the /tmp directory of a lambda that was being reused by AWS. I've temporarily solved this issue by emptying out the /tmp directory at the end of lambda_function.py just before the response is returned.
shutil.rmtree('/tmp', ignore_errors=True)