For the future you might want to consider whether an ini file with 85,000 sections is a good way of storing this.

For example a flat file with lines like:
a= (1). 1. f. Primera letra del abecedario espaņol y del orden latino internacional, que representa un fonema vocalico abierto y central. 2. f. Fil. Signo de la proposicion universal afirmativa. ~ por ~ y b por b. 1. loc. adv. punto por punto.a (2). (Del lat. ad). 1. prep. Precede a determinados complementos verbales, como el complemento indirecto y el complemento directo cuando este es de persona determinada o esta de algun modo personificado. 
a contrariis= (Loc. lat.; literalmente, 'por contrarios'). V. argumento a contrariis

could be read by $read(%fn,ns,a contrariss=) instead of $readini.

However this will still be a sequential read by mIRC of on average 1/2 of the file, which is likely to be resource intensive.

If memory is not a problem, you might want to consider reading the file into a hash table, and then saving / reloading the hash table.

Our you could use a database extension like "sqllite for mirc" (or whatever it is called) to access it more efficiently.