Improper parsing of text on Reuters - two paragraphs in the result instead of one

[ivanlabsii]

This is the sample page:
https://www.reuters.com/legal/qualcomm-saw-nuvia-buy-chance-save-14-billion-year-arm-fees-ceo-tells-jury-2024-12-18/

It has paragraphs like this:
<div data-testid="paragraph-0" class="text__text__1FZLe text__dark-grey__3Ml43 text__regular__2N1Xr text__small__1kGq2 body__full_width__ekUdw body__small_body__2vQyf article-body__paragraph__2-BtD">WILMINGTON, Delaware, Dec 18 (Reuters) - Internal Qualcomm <a data-testid="Link" target="_blank" href="https://www.reuters.com/markets/companies/QCOM.O" referrerpolicy="no-referrer-when-downgrade" rel="noopener" class="text__text__1FZLe text__inherit-color__3208F text__inherit-font__1Y8w3 text__inherit-size__1DZJi link__link__3Ji6W link__underline_default__2prE_ link__with-icon__3x3oD">(QCOM.O)<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 32 32" aria-hidden="true" focusable="false" role="presentation" data-testid="NewTabSymbol" class="link__new-tab-symbol__3T19s"><path fill="#666" d="M15 6c.6 0 1 .4 1 1s-.4 1-1 1H9c-.6 0-1 .4-1 1v14a1.1 1.1 0 0 0 .2.6c.1.2.4.4.8.4h14c.6 0 1-.4 1-1v-6c0-.6.4-1 1-1s1 .4 1 1v6a3 3 0 0 1-3 3H9a3 3 0 0 1-2.2-.9A3 3 0 0 1 6 23V9a3 3 0 0 1 3-3Zm10 0 .4.1.5.5.1.4v6c0 .6-.4 1-1 1s-1-.4-1-1V9.4l-5.3 5.3a1 1 0 0 1-.7.3 1 1 0 0 1-.7-.3 1 1 0 0 1 0-1.4L22.6 8H19c-.6 0-1-.4-1-1s.4-1 1-1Z"></path></svg><span style="border: 0px; clip: rect(0px, 0px, 0px, 0px); clip-path: inset(50%); height: 1px; margin: -1px; overflow: hidden; padding: 0px; position: absolute; width: 1px; white-space: nowrap;">, opens new tab</span></a> documents showed the chip firm estimated it could eventually save as much as $1.4 billion a year on payments to Arm <a data-testid="Link" target="_blank" href="https://www.reuters.com/markets/companies/O9Ty.F" referrerpolicy="no-referrer-when-downgrade" rel="noopener" class="text__text__1FZLe text__inherit-color__3208F text__inherit-font__1Y8w3 text__inherit-size__1DZJi link__link__3Ji6W link__underline_default__2prE_ link__with-icon__3x3oD">(O9Ty.F)<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 32 32" aria-hidden="true" focusable="false" role="presentation" data-testid="NewTabSymbol" class="link__new-tab-symbol__3T19s"><path fill="#666" d="M15 6c.6 0 1 .4 1 1s-.4 1-1 1H9c-.6 0-1 .4-1 1v14a1.1 1.1 0 0 0 .2.6c.1.2.4.4.8.4h14c.6 0 1-.4 1-1v-6c0-.6.4-1 1-1s1 .4 1 1v6a3 3 0 0 1-3 3H9a3 3 0 0 1-2.2-.9A3 3 0 0 1 6 23V9a3 3 0 0 1 3-3Zm10 0 .4.1.5.5.1.4v6c0 .6-.4 1-1 1s-1-.4-1-1V9.4l-5.3 5.3a1 1 0 0 1-.7.3 1 1 0 0 1-.7-.3 1 1 0 0 1 0-1.4L22.6 8H19c-.6 0-1-.4-1-1s.4-1 1-1Z"></path></svg><span style="border: 0px; clip: rect(0px, 0px, 0px, 0px); clip-path: inset(50%); height: 1px; margin: -1px; overflow: hidden; padding: 0px; position: absolute; width: 1px; white-space: nowrap;">, opens new tab</span></a>, by purchasing a little-known startup in 2021, according to evidence shown at a trial on Wednesday.</div>

This paragraph has this output:
<p>WILMINGTON, Delaware, Dec 18 (Reuters) - Internal Qualcomm </p><a data-testid=\"Link\" target=\"_blank\" href=\"https://www.reuters.com/markets/companies/QCOM.O\" referrerpolicy=\"no-referrer-when-downgrade\" rel=\"noopener\">(QCOM.O)<span>, opens new tab</span></a><p> documents showed the chip firm estimated it could eventually save as much as $1.4 billion a year on payments to Arm </p><a data-testid=\"Link\" target=\"_blank\" href=\"https://www.reuters.com/markets/companies/O9Ty.F\" referrerpolicy=\"no-referrer-when-downgrade\" rel=\"noopener\">(O9Ty.F)<span>, opens new tab</span></a><p>, by purchasing a little-known startup in 2021, according to evidence shown at a trial on Wednesday.</p>

As such it outputs two paragraphs, though the page contains just one.