If you have en ebook that split badly a paragraph, you can try (to merge the bad splitted paragraph), as find
</p>\n\n\s\s<p>([a-z])
so that a paragraph beginning with a lowercase word be merged with the previous one.
But in Calibre editor you must check “case sensitive”, otherwise [a-z] will be read as [A-Z], and all first letters (lower and upper case) will be selected.
And as replace you could try
\1
In this way you keep whatever letter is found with ([a-z]) in the result.