When a multipart form with file uploads is processed, parsed_body() (line 160 of _wsgi_common.py) sets file fields to AnnotatedValue.removed_because_raw_data() whose .value is "". The _default handler returns that empty string, so the span attribute ends up as {"file_field": ""} with no indication that a file was present but excluded.

The if "form" in request.__dict__: branch is never exercised by the new streaming tests — consider adding a test that POSTs application/x-www-form-urlencoded or multipart/form-data with span streaming enabled to verify form fields are captured on the segment span.

When max_request_body_size="never" is configured, request_body_within_bounds returns False (treating "never" the same as an oversized body), so the span attribute http.request.body.data is still written with "[Exceeds maximum size]". This contradicts user intent — they expect "never" to fully suppress body capture — and misattributes the reason: the body wasn't too large, it was explicitly opted out.

The if span_streaming: ... else: ... blocks throughout the test body create two essentially independent code paths with different fixtures, assertions, and even a conditional sentry_sdk.flush() call — consider splitting into two dedicated test functions for clarity and to avoid masking regressions in one path.