Problem

splitIncompleteMdx split fully balanced input like <a><b></b></a> into ['', '<a><b></b></a>'].

The parser reached the final > at end-of-string, incremented i, then immediately treated i >= len as “never found the closing angle.”

Fix

• Track whether the tag scan actually found >.
• Only mark the tag as incomplete when the scan exits without ever finding >.

Rule

When a small parser advances past a delimiter before checking end-of-input, keep an explicit foundDelimiter flag.

Otherwise balanced input at EOF gets misclassified as truncated.