See PHI1296. The "Download Raw Diff" flow generates a single diff blob, then turns it into an object in Files for download.

It does this in an older way that bypasses the chunk engine, and fails with a misleading error if the blob is larger than 4MB (and no engine is configured to store huge chunks, which they normally shouldn't be).

To reproduce:

• Create a 9MB diff.
• Configure one or more file storage engines in a reasonable way with an 8MB limit.
• Click Download Raw Diff.

Expected behavior:

• File is broken into chunks internally, stored, and downloaded normally.

Actual behavior:

• Failure with misleading "no configured storage engine can store this file" error.