For its time, perhaps. But "for its time" means "when there were exceedingly few TTRPGs and we knew almost nothing about designing them".
I don't agree with this notion that classic D&D's core design is poor or unsatisfactory for sandboxing.
There are wobbly parts to the game - for instance, how does a ranger's ability to track etc interact with the rules for getting lost? But the core idea, of the players saying where their PCs go; the GM then (subject to the getting lost roll) tracking their movement on a map; and the GM using a combination of random rolls and the map key to tell the players what their PCs discover/encounter/experience; in my view are pretty solid.
I think there are further questions that can be asked about the degree of player agency in this sort of "hexcrawl" play - I think that classic D&D relies, for player agency, on methods and approaches that don't work very well outside of a dungeon. But that doesn't mean that the resolution process, in itself, is not workable.
Narrowly true. But I'd consider a world with a large amount of fixed content and some random rolls a sandbox. The random rolls actually add to the sandbox, imo. And "roll encounters on table X" is a type of fixed content. So it's a fuzzy boundary.
I think random encounters are a pretty core part of the "classic" sandbox.
But they sit within a structure: they are
per <time period>; the time elapsed is calculated by reference to a pre-drawn map, and an already-established PC movement rate; the roll is made on a pre-authored table (as you say, a type of fixed content, and very analogous to a map key). So random encounters, done this way, reinforce the core resolution method, of map-and-key plus movement rates and movement tracking on a map.
They also involve improvisation/just-in-time authorship: the GM has to decide where the wandering beings comes from, what their motivations are, etc.
Reducing the reinforcement of the resolution method, while increasing the centrality of the improvisation, is one of the ways that post-classic-D&D design has built on classic techniques while doing slightly different things with them.
*****************************
For what it's worth, I have my own currently-favoured process for a sandbox-y feel to journeying, namely, Torchbearer 2e journeys.
This does use a map, but the map doesn't need a scale or even accurate proportions. All the map has to do is indicate landmarks. The number of settlements and/or landmarks the PCs are passing on their journey determines base
toll (normally 1 to 3 or 4-ish). Toll is also modified by weather (there are season-based random weather tables) and by a roll for "trouble on the road".
If the PCs aren't following a map, the players need to test Pathfinder (with the difficulty depending on distance, ie base toll). Adverse trouble on the road results also permit tests to negate toll penalties. And failed tests in TB2e permit the GM to narrate a twist - which is how "random" encounters, getting lost, etc happen in this system.
At the end of the journey, PCs suffer conditions (this is the system's general health/harm/debuff mechanic) equal to toll that they don't "buy off": toll is bought off via gear, supplies, spending money in settlements passed through, etc.
Like classic D&D, it uses a map (I use the World of Greyhawk for my TB2e game). Unlike classic D&D, the map can be shared with the players, as there is no secret component - eg if the PCs get lost, the GM does not need to secretly track movement. And there is no need to use movement rates or distances for resolution: these are just colour.
But I think the use of the map, and the calculation and buying off of toll, make it a bit more "gritty" and less rising action/climax-oriented than the Ironsworn approach.
