people keep calling it an 'onsite' but there is no onsite. everything is remote. 5-6 video interviews spread over one or two days depending on scheduling. just went through this for a senior SRE role (they call it 'reliability engineer' internally) so i can give a specific breakdown.
my final round schedule: hiring manager (already done pre-final round) technical interview 1: infrastructure/systems design (60 min) technical interview 2: coding in CoderPad (60 min) technical interview 3: another systems design, different interviewer (60 min) behavioral 1: values/collaboration (45 min) behavioral 2: 'leadership and results' (45 min) 'coffee chat' with potential teammate (30 min, explicitly not evaluative they said -- i'm skeptical)
so yeah. long day. i spread it over two days and would absolutely recommend that. doing all six in one day is brutal.
what surprised me: the interviewers all knew each other and had clearly looked at my submission materials. no 'so tell me about yourself' cold opens. they'd start with something specific, like 'i saw you've worked with GitLab runners before, let's build on that.' made it feel efficient.
the 'not evaluative' coffee chat: the person i talked to was genuinely casual, we talked about the team's on-call rotation and what it's like to work fully async with teammates in India and Ireland. i don't know if that gets fed back to the hiring committee but it's definitely your chance to ask honest questions about the job reality. i asked about pager volume and got a real answer.
post-round timeline: i had the full round on Tuesday/Wednesday. got a call from the recruiter Friday afternoon saying they were extending an offer. genuinely fast. the offer call was separate from the actual offer letter, which came Monday morning.
debrief/feedback: they don't give structured written feedback if you decline or get rejected. the recruiter said she'd share 'high level themes' if i asked, but nothing specific. standard tech company behavior.
if you're going in for SRE or infrastructure: know your incident response process cold. they asked me to walk through a specific production incident start to finish in both tech rounds. the reliability angle is central.