The vulnerability stems from improper buffer management in HttpRequestConduit.java's processWrite method. The pre-patch code in STATE_START (lines 126-141) used assertions to validate buffer capacity but lacked proper handling for oversized URLs. The patch introduced a length check (line 130-133) and a fallback path (lines 145-156) to write the URL incrementally via STATE_URL. The unpatched code's direct buffer writes without overflow checks allowed BufferOverflowExceptions when URLs exceeded buffer limits, triggering an uncaught exception loop. The HpackEncoder.java changes addressed secondary overflow scenarios, but the root cause was in HttpRequestConduit's request line construction logic.