Only consider it a block waiver if all future stops are SKIPPED

[lemald]

Summary of changes

Asana Ticket: 🧠 Trips are only cancelled when all of their StopTimes are cancelled

This is presently running on dev-blue.

On prod, we see the following when trying to plan a trip for later in the day on the 71:

However, on dev-blue, we get the correct trip plan:

Meanwhile, genuine cases of block waivers continue to receive cancellations:

[paulswartz]

I didn't dig into it too much, but it does look like this is adding about 500ms processing latency: https://mbta.splunkcloud.com/en-US/app/search/search?earliest=%40d&latest=now&q=search%20index%3Dconcentrate-dev*%20update_latency_ms%7C%20eval%20update_latency%20%3D%20(update_latency_ms%20%2F%201000)%20%7C%20timechart%20median(update_latency)%20by%20index&display.page.search.mode=fast&dispatch.sample_ratio=1&display.general.type=visualizations&workload_pool=&display.page.search.tab=visualizations&sid=1760723853.21367

[lemald]

I didn't dig into it too much, but it does look like this is adding about 500ms processing latency: https://mbta.splunkcloud.com/en-US/app/search/search?earliest=%40d&latest=now&q=search%20index%3Dconcentrate-dev*%20update_latency_ms%7C%20eval%20update_latency%20%3D%20(update_latency_ms%20%2F%201000)%20%7C%20timechart%20median(update_latency)%20by%20index&display.page.search.mode=fast&dispatch.sample_ratio=1&display.general.type=visualizations&workload_pool=&display.page.search.tab=visualizations&sid=1760723853.21367

Interesting, thanks for pointing that out. I tried to make it so that this code path runs only if absolutely necessary to keep processing time down but it seems like that isn't enough. I'll deploy your MapSet suggestion and see if that helps at all.

[lemald]

Okay, the MapSet thing didn't have any noticeable improvement to performance so I've tried some fiddling around with the ETS tables so that we can get all of the information we need with one query rather than querying the table in a loop. We'll see if that's any better.

[lemald]

Update: processing time is down, but confusingly it's down across al dev environments and starting before I deployed the change. I'll let it run longer so that we can see if any trends emerge.