Or, to put that another way: there's no real difference between a bug introduced by wrapping a game in an imperfect emulator, and a bug introduced during recompilation/porting/remastering of the game. Either way, you now have a new "variant" of the game with its own bugs, but one which is also a canonical, supported release of the game.
Interestingly, sometimes—because of one of these slight variances—the fastest [and so preferred] version of a game to speedrun, is a Nintendo-sanctioned emulated (e.g. Virtual Console) release of the game. It says something important, I think, that these releases of the game aren't automatically shunned by the speedrunning community, the way that runs of the game under an arbitrary emulator on someone's PC would be; but rather are just treated as their own separate category-set, in the same way the speedrunning community differentiates different region releases, or version releases, or port/remaster releases.
Not that that necessarily has anything to do with bugs. Some VC games simply have reduced lag when a lot is loaded, because they throw somewhat more power at the game than the original console could.
One more interesting difference, that you'll see with N64 games specifically (e.g. VC Paper Mario on the Wii) is that the original console will hard-lock in places where the VC version will just keep on truckin' (presumably because it has access to more memory, or the emulator just discards crazy out-of-bounds writes, or because the emulator treats open-bus address reads differently. Something like that.)
Anyway, that means that there are glitches that can be used save time on the VC version, which in some sense don't exist at all on the original version of the game, because they're gated behind a fault that the original hardware can't get past.
You'll see the effect of this in gaming in the forms of e.g.:
• The https://www.no-intro.org project, whose goal is to curate a database of hashes of "canonical ROMs" (i.e. ones that are don't have demoscene "intro" greetz; but also which don't have the extra headers that some copiers add; and, obviously, which aren't bad dumps.)
• The policies of some emulators (e.g. MAME, Higan) to replace, whenever possible, High-Level Emulation code that "works perfectly" (i.e. matches console output on every frame for every known software title!) with Low-Level Emulation, which requires a BIOS/firmware dump of a chip to run. The goal in these types of fixes isn't to run existing games, but rather to make the emulator match the behavior of the hardware for running even novel or previously undiscovered games, such that it can be a useful tool in future conservation efforts, when the original hardware is long gone.
• The file-formats of some emulators (again: MAME, Higan) which keep individually-dumped coprocessor ROMs as separate files in a software title's archive file/folder, each with their own metadata; rather than inventing custom merged-binary file-formats for each system. This both allows future archivists to easily re-verify that individual dumps are correct; to replace bad dumps with better ones without invalidating the hashes for other ROMs that make up the same software title; and, in theory, this provides a (coincidentally machine-executable) schematic of the software-title-as-hardware-system, which could be used to reconstruct it wholesale in a future where all existing hardware was lost.
• The transition from the archivists at Archive.org preferring to see Apple II disk images in raw bit-dump (DSK) file-format, to wanting them as flux-trace (https://applesaucefdc.com/woz/) files, which in turn means that copy-protected games can now be preserved in their original, copy-protected state, rather than first needing to be cracked.
If the publisher modifies the work particularly in a manner devoid of creative intent—such as unintentionally omitting page 527—the modification cannot possibly be considered canonical. Imperfections in an emulator are not the work of the original author and they're certainly not any form of artistic intent.
https://arstechnica.com/gaming/2018/11/sony-using-open-sourc...
I believe the consensus is that the NES classics on GBA were done using an open source emulator. The emulator itself was liberally licensed and I think someone asked the author's thoughts and they were 'ya, that's half the purpose of the license. i don't care'