Cool game! I think the general concept has promise but it needs a lot of refinement.
The general problem is that runs feel too much the same, however I think you already know this because the content you intend on adding to the full game is targeted towards fixing this.
Even so I think the way that the current "memory" system contributes a lot to the problem. Most of the memories are strong and strictly positive, and you get a lot of memories; so many that it can be difficult to keep track of them all. To me the problem can be split into two parts;
First, the builds you create over the course of different runs don't feel very unique or "defined". To illustrate what I mean by the term "defined"; a run in a deckbuilder like balatro or StS is very defined in the sense that I can describe to a player the "idea" of the deck in terms of its main combos, mechanics and archetypes without telling them much about the specific cards or relics/jokers. The existence of defined builds implies the ability of the player to coordinate their choices so as to create cohesion or purpose within the build, which is one of the things that people enjoy in these sort of games.
In general the appeal of a roguelike mostly falls into the inherent appeal of the basic gameplay stuffing (so like the omens here), as well the overarching "run building" decision making. This decision making itself sits on a spectrum from local/micro decisions of the form "how do I influence my build so as to do well in the immediately upcoming encounters", and the more global/macro decisions about how to build something cohesive; "how do I build powerful synergies".
Although not especially deep at this point in time, the game does have some micro decision making; I was thinking about which order to take the omens in given how hard I was finding them/how much easier they would get, and I could often choose omens specifically to target the difficulties I anticipated based on the remaining omens.
On the other hand run definition lies in the macro decisions and this game really doesn't have any of that; instead each run is defined by an uncorrelated mush of something like twenty different memories.
I think it's ok for this game to not lean too heavily into synergistic deckbuilding because not everything has to be balatro, but I also think that it would be clearly better if the player had a greater ability to define their run. If you want to make this thing the best version of itself you should probably experiment with a lot of different versions of the run building system. I don't claim to know what's optimal here, but a simple idea that springs to mind is having a limited number of slots for memories (like jokers in balatro) which you can swap out, which I think adds a bit more depth and choice to the whole affair (a cool thing here is that there is a strong incentive to swap out elements of your build as the omens you focus on change). Relatedly, it limits the number of memories to the point where you can think about each of them individually and the run has a concrete identity. Also the adding sand option isn't very pickable right now (which is part of why you get so many memories), and I think it might be interesting if you got sand for skipping a choice of memory after seeing the options rather than having to commit to it before seeing what you might get (I'm not actually so sure about this one, but the point is that these are all things to consider).
Looping back, the second problem is that the memories disproportionately affect certain games compared to others. The shooter games end up being fairly trivial if you leave them to the end, because you will be filling half the screen with randomly-size explosive splitting bullets, whereas the voyager game would probably be more difficult in later stages than at the start of the game (the only thing which seems to help here is some aura collection effects and enemy slowdown). There is also the fact that the games are generally a little unbalanced; in particular I think the bubbles game is really easy (maybe would be more interesting with some kind of time pressure?). To some extent the fact that certain omens are better to play at different stages of the run adds strategy, but in this case the strategy is not very subtle or interesting.
All these issues can be addressed by balancing the games and memory pool with data from playtesters. On the other hand every player will have a different experience with the relative difficulty of the games and so another consideration is to design the game in a way which is innately resilient to these differences. Namely the game should probably allow the player to manage the relative difficulty of the games with their decision-making. I think this consideration is kind of baked into the premise of the game, and so maybe this problem just goes away if the run definition is improved.
I will cap the unsolicited advice here. Looking forward to the full release!