[16:21:26] <ori>	 TimStarling, Krinkle: can you help me reason through Excimer sampling? I have two questions:
[16:21:54] <ori>	 (a) Is there a straightforward proportional relationship between the profiling period and sample rate? There's now a separate log sink for PHP 8, but it's only gotten one sample so far due to limited rollout. Would cutting the profiling period from 60s to 10s for PHP 8 yield approx. 6x more samples?
[16:22:17] <ori>	 (b) Would comparing aggregate data from PHP 7 (sampled with a 60s period) and PHP 8 (sampled with a 10s period) be a valid way to identify regressed or improved symbols, or would the different sampling rates skew the comparison by also altering *which* requests are being sampled?
[16:23:01] <ori>	 cc cdanis 
[16:51:11] <ori>	 oh, it's already saturday there. stupid timezones.
[17:41:05] <bd808>	 ori: possibly related, PHP 8.1 got rolled back because of performance concerns in T385395
[17:41:05] <stashbot>	 T385395: 503 error when edit large size pages on PHP 8.1 - https://phabricator.wikimedia.org/T385395
[17:41:23] <bd808>	 so samples would have stopped
[19:01:09] <ori>	 bd808: OIC. thanks
[23:52:02] <TimStarling>	 ori: it was 3am Saturday
[23:53:15] <TimStarling>	 a) yes b) yes (to the former proposition)