As I said yesterday, various words relating to time have two forms: One for past and one for future events. These words are an example. They are constructed of sji which is found also in other words which start with ‘every-‘ in English and the past or future tense marker. The words have of course no tense seeing that they are no state verbs and the position of la or ki is not anyhow like in their meaning as tense markers (where they are prefixes between number and subjunctive prefix), but rejistanian prefixes do work in several jobs to make ends meet. It was a design decision to re-use suffixes when it could at least vaguely be justified, because words like kiam, iam cxiam in Esperanto annoyed me. They just refered to nothing. If Zamenhof used the -as of the present tense there instead, it would have made much more sense. So I adopted the idea into rejistanian, only that the (unmarked) present tense could not be used.
A good way to disambiguate between these 2 words would be the sentences Min’la’va jilih reja jilih sjila (We-PST-do this method this always: We have always done it like that) and Min’ni ‘va jilih reja jilih sjiki (We-must (INF)do this method this always: We must always do it like that).
Other words which change between la and ki forms are lata/kita (never), lajitax/kijitax (later), lystas/kijistas (earlier*), sunla/sunki (when), ulja/ulki (sometimes*) and ulsila/ulsiki (which is similar to the previous word but with a higher degree of uncertainty or a lower degree of caring).
* the la form here changes due to its company but it is there.