Stuck in "got prompt" when using large workflows

[jeffersoncgo]

I have a really large Workflow, wich various custom nodes, and it was working really well, until i add "to much loras", it's a custom lora script, wich has the option to bypass it as parameter. I redid the script using the core LoaderLora as base, to try to avoid this problem, but didn't work.

import os
import torch
import comfy
import folder_paths
from nodes import LoraLoader

class Lora_Loader(LoraLoader):
    @classmethod
    def INPUT_TYPES(s):
        types = super().INPUT_TYPES()
        types["required"]["lora_name"] = (folder_paths.get_filename_list("loras"),)
        types["required"]["bypass"] = ("BOOLEAN", {"default": False},)
        types["optional"] = {}
        types["optional"]["triggers"] = ("STRING", {"default": "", "multiline": True},)
        return types

    RETURN_TYPES = ("MODEL", "CLIP", "STRING", "BOOLEAN")
    RETURN_NAMES = ("MODEL", "CLIP", "LORA_NAME", "BYPASSED")
    CATEGORY = "JeffisNotHere / Loaders"

    def load_lora(self, **kwargs):
        if kwargs["bypass"]:
            kwargs["strength_model"] = 0 #Before i was just returning it here, but changed to this to still go through the core script, to see if it worked, (it didn't)
            kwargs["strength_clip"] = 0
        model_lora, clip_lora = super().load_lora(**kwargs)
        return (model_lora, clip_lora, kwargs["lora_name"], kwargs["bypass"])

NODE_CLASS_MAPPINGS = {
    "Jeff - Lora Loader": Lora_Loader,
}

If i try to load row by row, it works (until reach the stuck zone), but if i already in the begin try to load from the Slow zone, it also become stuck.

All of them are using it's own Bypass param as true, so the loras don't load, even so, still getting stuck, is this related to limitations in data json size/lenght in the post method?

In the print i send just a part of the Workflow, that is the part bugging.

[jeffersoncgo]

Oh, i almost forgot, it become stuck in here:

No matter how much time it takes, it don't go beyond the "got prompt" message.

I edited the "execution.py" file, and inserted "print" messages in each function to see where it was stuck.

And it sometimes get stuck endelessly in recursive_output_delete_if_changed and sometimes in recursive_will_execute

[poisenbery]

Bumping this. I have an extremely complex workflow as well, generation will hang for about 15 minutes before actually doing anything.

I removed about 30 seconds of processing nodes, and the entire generation was shortened by 14 minutes.