First and foremost, the bold "A" is a mistake. The Perseus Project digitized the lexicon mostly into a plain-text form, which is then re-processed to turn citations like "Il.20.297" (Iliad book 20 line 297) into links, add in nice-looking headers for readability, and so on.
This processing isn't always perfect: look at "ἔπα^θον" at the end of the first complete line, which is meant to be ἔπᾰθον (with a breve on the alpha). And in this case, it also inserted a header somewhere it didn't belong. That first line is supposed to start with impf. ἔπασχον 17.735, etc. : fut. πείσομαι Od.2.134, etc.—in other words, listing out different forms a student might need to conjugate it properly in different tenses, with pointers to where those forms are attested.
After the tense information, it launches into the first translation, which is where there should potentially be a paragraph break. This starts at the end of the third full line, with "have something done to one, suffer". The parts in italics in each translation are the ones corresponding to the word, so the formatting here is meant to show that the "something" isn't actually part of πάσχω itself (it would be represented by a separate word or phrase in the Greek).
Definitions are then organized hierarchically, which is where the Roman numerals, Arabic numerals, and lowercase letters come into play. As you correctly surmised, all the numberings start with 2; the intent is that you can cite the meaning directly under II as "definition II.1", without ever needing to print two heading numbers in a row. There should be a I before the first definition, but it's missing here, which I think is a processing issue.
Within each definition, there are some attestations with that particular sense. The actual word being defined is abbreviated to its first letter to save space, so when you see κακῶς π. ὑπό τινος, that π. is some form of πάσχω. (The particular form generally doesn't matter without more context.)