Now, to my understanding, the first part of this is exactly how the original fantasy RPG began play under Dave Arneson -- players took the parts of Heroes (worth 4 figures) from Chainmail, and advanced them to Superheroes (worth 8 figures) through play. In those rules, your basic Hero did in fact make 4 attacks for each normal man figure.
Sometimes we lose sight of that fact, since it's unstated in the OD&D books, which presume we're referring back to the Chainmail man-to-man combat system. But the point was reinforced in the article "Questions Most Frequently Asked About Dungeons & Dragons Rules", from The Strategic Review #2, p. 3 (Summer 1975, Editor: E. Gary Gygax):
Combat Example: 10 ORCS surprise a lone Hero wandering lost in the dungeons... Note that he is allowed one attack for each of his combat levels as the ratio of one Orc vs. the Hero is 1:4, so this is treated as normal (non-fantastic) melee, as is any combat where the score of one side is a base 1 hit die or less. Hero: 19; 01; 16; 09. Two out of four blows struck...
Now, it's made clear in this passage -- and again elsewhere -- that the multiple-attack rule is here being limited to opponents who have 1 hit die or less (and that's why you see the rule appearing in AD&D the same way). Nevertheless, it's clear that the game assumes that the PCs are facing down very large numbers of normal opponents, and hewing through them quite rapidly (even in a 1:10 ratio, as seen here). Consider also Arneson's "Temple of the Frog" in the Blackmoor supplement, where the first level of the dungeon contains barracks housing 50, 100, 250 men each, etc.
The point is, the original play of the OD&D game assumed that your Fighters were cleaving through enormous numbers of non-fantastic enemies, and obviously the Wizard-types had to have something to keep up with them. The sleep spell as we see it in OD&D exactly matches this situation -- 2-16 normal men affected (9 on average), a bit more than the number of attacks that a Hero would dish out in 2 rounds. (And you can also see why Greyhawk stipulated the first no-save rule for sleep; the number of saves you'd have to roll would otherwise be enormous.)
So if we switch to a system where Fighters are not getting these Chainmail-esque machine-gun attacks, then this wouldn't make sense anymore. Certainly, I play with everyone just getting one attack all the time. If I were to read the OD&D books by themselves, that's certainly how I would interpret the combat system. And after playing 3E for a number of years, I'm solidly in the camp that the multiple-iterative-attacks mechanic there was really a huge mistake. Let's proceed with the more elegant rule of one basic attack resolution for all PC-types.
If our OD&D Heroes only get one attack against normal foes (1:4 original), then our sleep spell must also be reduced in the same ratio to remain balanced with it. Say, take the basic effect (2-16, average 9), divide it by 4 (9/4 = 2.25), and find a reasonable die for that range (d4, perhaps d6 if we prefer the shape?). And do the same thing for the other hit die categories (2HD: average 7/4 = 1.75, say 1/2 d6; 3HD: average 3.5/4 = 0.88, say just 1 creature; 4HD: average 1/4 = 0.25, say no creatures affected). We do this while acknowledging that the D&D Fighter, under the TSR #2 rule, doesn't get multiple attacks against higher-level foes (it would be very odd to give sleep a greater number of upper-HD affected).
So that's what I think I'll do for my game's sleep spell. We'll have sleep affect 1d6 1HD, 1/2 d6 2HD, or 1 3HD figures (and none of higher level). That's in harmony with the reduced number of Fighter attacks and presumed number of enemies, as compared to Chainmail. It seems about fair to me, and in the same spirit as the original environment for the spell.
Finally, here's some other comments. There's no need to impose Greyhawk's no-save rule here, so I'll continue to afford saves to the targets of the spell. Flavor-wise, I feel that sleep-enchantments should be generally more difficult to break out of than normal sleep. Therefore, I see no reason for more than a 2-in-6 chance to wake up a victim by way of shaking/slapping, etc. It should also last a fairly long period of time (duration is unstated in OD&D), perhaps a full sleep cycle -- I'll say 12 turns (a common duration in Vol. I), although if you said 12 hours I'd be quick to agree with that, as well.