`clerk.users.count` with arguments can be very slow for large number of users

Repro

import time
import clerk_backend_api
from clerk_backend_api import Clerk

print(clerk_backend_api.__version__) # 3.1.11 (latest)

clerk_api_key = "xxxxx"
clerk = Clerk(bearer_auth=clerk_api_key)

start_all = time.perf_counter()
all_users = clerk.users.count().total_count
end_all = time.perf_counter()

print(f"{all_users = } (took {end_all - start_all:.3f} seconds)")
# all_users = 926354 (took 0.342 seconds)

created_before_ms = int(time.time() * 1000)

start_before = time.perf_counter()
users_before_ts = clerk.users.count(created_at_before=created_before_ms).total_count
end_before = time.perf_counter()

print(f"{users_before_ts = } (took {end_before - start_before:.3f} seconds)")
# users_before_ts = 926354 (took 436.951 seconds)

As you can see, adding a predicate to users.count changes the response time from sub 1 second to ~7 minutes.

I appreciate the "COUNT *" fast path, but 7 minutes feels very long in the other case.
Is it possible that performance can be improved here at all?

Start with the provided Python reproduction and the clerk.users.count entry point, comparing the unfiltered call with the created_at_before argument. Trace how the argumented count request is handled and define done as a substantially shorter response time for large user sets without changing the count result.