Or it can work.

Not both

> which disabled people rely on

is pretty clear implication. But I'll spell it out for you - people with impairment to senses (primarily vision) rely on this infrastructure to know when it's safe to cross a road.

---

It can be simultaneously true that "this particular incident resulted in no actual harm (because the prank messages still usefully conveyed the required information)" and "messing with accessibility devices has the potential to cause harm, and it would be preferable for well-intentioned pranksters to direct their efforts elsewhere".

In particular, I'd still push back on a claim that these messages were still useful because they conveyed the same instruction. Someone relying solely on this announcement (without the ability to visually verify the situation) might have reasonable cause to doubt whether the announcement is still trustworthy. After all, they might reason, if a prankster is able to change the message, might they also able to change the playback conditions? Or might they be able to switch the "don't walk" and "walk" announcements? Anything which causes the UX to deviate from the known, reliable, trustworthy pattern introduces significant trust-costs.

Did you say thank you today? - A message from VP JD Vance

Reminder: Vandalism of Tesla vehicles or properties is an act of terrorism and is punishable by deportation

Yes, emotional maturity can help, and it also depends on perspective (whether I identify with the author of the PSA). But when the situation gets more complex than picking up litter, I think people forget that this dynamic exists and is important.