Lucene nightly benchmark regression on 29/01

### Description

We are seeing regressions on lucene nightly benchmarks on [29/01](https://benchmarks.mikemccandless.com/2026.01.29.14.59.00.html)

The set of commits which seems to be responsible between [22/01](https://benchmarks.mikemccandless.com/2026.01.22.18.03.42.html) (the last run) and [29/01](https://benchmarks.mikemccandless.com/2026.01.29.14.59.00.html) seems to be these: https://github.com/apache/lucene/compare/2f9aa8ae26d6c1087884c734e1b3d137bd8c6601...6a6b753e3080725921a07b0a214963d8ff639eea.

We should find the root cause of the regression.

### Version and environment details

_No response_

Start with the linked nightly benchmark reports for 2026.01.22 and 2026.01.29, then inspect the commits in the provided comparison range. Reproduce or isolate the regression using the Lucene nightly benchmark process and identify the responsible change; done means the root cause is documented and the regression is addressed or clearly handed off.

Written by the indexing model from the issue text.