Vote #1 Christopher Kowalewski: That’s the working through of internal reconstruction that you only see the results of in the textbooks.
Now, Chad Turner suggests I’d know the answer. God, *I* don’t know the answer. But Sihler does: New Comparative Grammar of Greek and Latin. Or at least, Sihler knows as much of the answer as there is to be known.
§531. Other tense and mood stems of the Latin perfect system are all based upon a combination of the Latin perfect stem, whichever that happens to be, with an element *-is-, of wholly obscure origin but most commonly imagined to be related somehow to the s-aorist. […]
1. Pluperfect Indicative. Functionally the past of the perfect, in effect, like New English had gone. Latin -eram, -erās, -erat from *-is- together with *-ā-, the optative formation which functions as an anterior tense marker in -bā- and erā- ‘was’. (The vowel of -er- is of course ambiguous per se, but its historical value can be surmised from the pluperfect in *-issē-, 4, below.) […]
3. Perfect Subjunctive. […] So *-is-ī- plus endings, whence by regular sound laws –erim, -erīs, -erit, -erīmus.
4. Pluperfect Subjunctive. Descriptively, the stem in *-is- with an additional element *-sē- of profoundly obscure origin; so *-is-sē-, whence -issem, -issēs, -isset.—The same *-sē- is found in the imperfect subjunctive.
So, to unpack this. If we look at the pluperfect ama-v-eram ‘I had loved’, the pluperfect person inflection (-m) comes from the same place as the person inflection of eram ‘I was’. And the vowel -ā- in -eram comes from the same place as the -ā- of eram. But the -er- in the suffix is not the same stem as the er- in eram. It is a development of proto-Latin *-isām.
And how do we know that the “wholly obscure” *-is- infix is not the same as the stem of esse? Because the -i- still shows up, e.g. in the pluperfect subjunctive. But also because a verb doesn’t get just plunked at the end of a finite verb, †amavi eram. If you have a compound verb in Latin, you expect the auxiliary to be combined with a non-finite verb, like amare habeo > French aimerai.
Add that the perfect –v– is specific to Latin (it’s absent from Sabine: Sihler §528), so it can’t be that old in the language: certainly not old enough that an †amavXXXXX eram could have been lost in the depths of time.