1. 1. Motivation

`backup_restore_stop_start.js` was timing out on the `enterprise-amazon-linux2023-arm64-all-feature-flags-with-replicated-size-and-count` variant (BF-43460, recurring ~4× in 30 days). That variant enables `featureFlagReplicatedFastCount`, which writes to internal fast-count tracking collections on every user write — roughly doubling the effective oplog volume. The hidden node added at the end of the test has to replay all that oplog, and with the extra overhead it couldn't catch up within the implicit timeouts, causing `rst.awaitReplication()` to hang silently until resmoke's hang analyzer SIGKILLed the shell.

1. 1. What this PR does

Removes the explicit `clientTime: 30000` from `backup_restore_stop_start.js`, letting it fall back to the library default of 10 seconds. This reduces oplog generation by ~3×, giving the hidden node substantially less to replay.

`backup_restore_rolling.js` legitimately keeps its 30-second `clientTime` because it performs three rsync passes with 10-second sleeps between them and needs data to sync across those passes. The stop-start backup is instantaneous (stop → copy → start), so the extra 20 seconds only generates more catch-up work without adding meaningful coverage.

1. 1. Reviewer reading guide

Only one line changes: `backup_restore_stop_start.js:22`.