The Lightning Network does not scale.

This is why everyone is building fake centralized hybrid Lightning bullshit and using custodial apps.

There are many reasons for this, here are some.

Liquidity - Each channel requires funding, twice. That capital is frozen until you splice, rebalance, or close. As adoption grows, the liquidity requirement grows quadratically. As on-chain fees grow, cost of using LN goes up while security assurances go down.

Complexity - Complexity breeds centralization. most users do not tolerate the base LN experience. Most devs do not tolerate depending on complex buggy self-custodial implementations. You've practically got to be an entire Lightning stack business to reliably provide a self-custodial product. Costs rise, leading to severe centralization at scale.

Routing - Portrayed as the coolest part of LN, but truly the worst aspect due to liquidity requirements, uncertainty, complexity. Results in hubs, then centralization at scale.

Breaking changes - Constant new complexity requires node runners to always run new, potentially insecure software. New channel types, new payment protocols, all destroy interoperability.

Obscure Hacks Required - If you want to provide a LN wallet or app you need to learn all the weird solutions, like LNURL, misc patches & tools, that people hacked in because LN protocol devs and LN implementation companies rarely care about the user space (probably because it is hopeless). We get weird derivative things hacked into others, like subscriptions into one specific payment protocol, but not into others; or, weird email format nicknames that arent actually emails, and are all implemented in trusted ways. This results in LN businesses and LN devs requiring arcane understanding, and endless patience, in the LN world.

Regulatory trap - Running a LSP business safely requires experienced lawyers for a constantly changing compliance landscape. New hybrid LN services like Spark, Liquid swaps, and taproot-asset edge nodes, will draw regulatory scrutiny the moment something goes wrong, or becomes too large.

Lightning is still cool and useful, but it doesn't actually fix Bitcoin payments at scale, so much as kinda-sorta provide efficiencies as long as you don't actually scale too much...

TLDR?

The concept of a high-frequency bitcoin channel is sound, and proven now.

The concept of a bitcoin-based routing network as an efficiency has not been proven, and, arguably, has failed.